LONG-TERM EFFECTS OF TIMBER HARVESTING ON COARSE WOODY DEBRIS DYNAMICS AND MARTEN HABITAT by Ingrid Farnell B.Sc., McMaster University, 2015 THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in NATURAL RESOURCES AND ENVIRONMENTAL STUDIES UNIVERSITY OF NORTHERN BRITISH COLUMBIA April 2021 © Ingrid Farnell, 2021 Table of contents Abstract .......................................................................................................................................... iv Acknowledgements ..........................................................................................................................v 1. Introduction ..................................................................................................................................1 2. Chapter 1: The effects of variable retention forestry on coarse woody debris dynamics and concomitant impacts on American marten habitat after 27 years ....................................................5 2.1. Introduction ..........................................................................................................................6 2.2. Methods .................................................................................................................................9 2.2.1. Study area .......................................................................................................................9 2.2.2. Experimental design .....................................................................................................10 2.2.3. Data collection ..............................................................................................................11 2.2.4. Data analysis .................................................................................................................13 2.3. Results ................................................................................................................................17 2.3.1. Harvesting effects on CWD volume through time .......................................................17 2.3.2. Harvesting effects on CWD decay class through time .................................................19 2.3.3. Windthrown CWD inputs through time .......................................................................20 2.3.4. Harvesting effects on CWD attributes, and the marten CWD habitat index, 26 years post harvest .............................................................................................................................21 2.3.5. Variation in CWD across forest types ..........................................................................23 2.4. Discussion ..........................................................................................................................25 3. Chapter 2: Area-based approach for modeling coarse woody debris characteristics using ALSderived measures of forest structure ..............................................................................................30 3.1. Introduction .........................................................................................................................30 3.2. Methods ...............................................................................................................................35 3.2.1. Study area .....................................................................................................................35 ii 3.2.2. Experimental design .....................................................................................................36 3.2.3. Data acquisition ............................................................................................................36 3.2.4. Linking empirical and ALS data ..................................................................................38 3.2.5. Landscape predictions ..................................................................................................40 3.3. Results .................................................................................................................................41 3.3.1. ALS CWD detection.....................................................................................................41 3.3.2. Landscape-level ALS treatment effect detection .........................................................42 3.4. Discussion ...........................................................................................................................46 4. Chapter 3: Conclusions ..............................................................................................................52 4.1. Overview ............................................................................................................................52 4.2. Key findings .......................................................................................................................52 4.3. Implications ........................................................................................................................53 4.4. Limitations and future research ..........................................................................................54 Appendix 1 .....................................................................................................................................56 Appendix 2 .....................................................................................................................................58 References ......................................................................................................................................61 iii Abstract At a long-term experimental trial in northern British Columbia, Canada, I analysed the impact of varying amounts of timber retention: 0% (clearcut), 40%, 70%, 100% (unharvested) on CWD volume, decay class, and inputs from windthrow over 27 years. I used attributes (diameter, length, decay class, and height above the ground) known to be favourable to marten to create an index for assessing the impact of harvesting intensity on CWD habitat features. I then used airborne laser scanning (ALS) to predict CWD volume, habitat value, and piece attributes over the landscape. Stands with 70% retention had CWD attributes that resulted in CWD habitat features similar to unharvested stands. Clearcuts contained pieces that were smaller, more decayed, and closer to the ground, which contributed less valuable habitat, compared to stands with higher retention. Windthrown trees were the majority of CWD inputs and volume change was positively related to percent retention. iv Acknowledgements I would like to thank the B.C. Ministry of Forests, Lands and Natural Resources and an NSERC Discovery grant for their funding. Thank you to the original team of researchers who initiated the Date Creek Study: Dave Coates, Phil LePage, Allen Banner, and especially Doug Steventon who led the CWD sampling effort. It has been an honour to continue the research of CWD at Date Creek. I appreciate the many individuals who assisted with the field data collection: Teresa White, Ann Macadam, Todd Mahon, James Cuell, Dave Patterson, Jeffrey Lemieux, Kristi Iverson, Luizmar De Assis Barros, Rafael Camargo, Juan Pablo Ramires Delgado, Caren Dymond, Kristen Hirsh-Pearson, Miguel Arias Patino, Rajeev Pillay, Oscar Venter, and especially my field crew Carleen Paltzat, Alicja Muir, Leonard Bannier, and Thibault Roussel. The Date Creek experiment is within the Gitxsan Laxwiiyip (house territories) of Xsa Gay Laaxan (Simgiigyet Mauus, Bill Blackwater Jr.) and Tsihl Hlii Din (Simgiigyet Xhliiyemlaka, John Olson). To my supervisory committee, thank you for taking the time out of your busy schedules to provide me feedback that has helped refine the final product. Erica and Anne-Marie, you have been incredible female mentors, and your ecological insight has been crucial in applying my research to forest management. A big thank you to Dr. Che Elkin, your patience, guidance, and continued support has been crucial in the completion of this thesis. Thank you to the Mixed Wood Ecology lab members for being a great team, sharing your problems, solutions, and experiences. Lasty to my partner Jeff, thank you for your patience (especially as I went over the two years), being a sounding board, and especially for your unwavering support. v 1 1. Introduction 2 3 Forest ecosystems are composed of many different components including live trees, 4 understory vegetation, non-vascular plants, insects, and wildlife. One component that is often 5 overlooked is fallen dead wood, also referred to as coarse woody debris (CWD). There are 6 challenges surrounding the benefits and hazards that CWD creates, often dividing forest mangers 7 and ecologists. As early as the 1920’s dead wood was identified as being an important ecological 8 unit by Graham (1925), however, as intensive commercial exploitation increased with the 9 industrial revolution CWD was considered a waste of economic resources, a source for pests and 10 disease, and a safety hazard. Resultantly, it was often completely removed from site during 11 typical timber harvesting in many countries. After Harmon et al. (1986) published a review of its 12 ecological function, CWD was increasingly studied by ecologists (Jia-bing et al. 2005). Early 13 studies focused on CWD’s influence on nutrient cycling (Fahey et al. 1988; Arthur and Fahey 14 1990; Spears et al. 2003), decomposition (Graham and Cromack 1982; Edmonds, Robert L. et al. 15 1986; Harmon, Mark E. et al. 1987; Stone et al. 1998), forest succession (Sollins et al. 1987; 16 Franklin et al. 1987; Spies and Cline 1988; Sturtevant et al. 1997), habitat (Harmon and Franklin 17 1989; Bull et al. 1997; Bowman et al. 2000), and biodiversity (Harmon et al. 1986; Maser et al. 18 1988; McMinn and Crossley 1996). In the 1990’s studies on the effects of forest management on 19 the quantity and quality of CWD and associated saproxylic-species (dead-wood-dependent 20 species) increased (Hansen et al. 1991; Buskirk and Ruggiero 1994; McCarthy and Bailey 1994; 21 Sturtevant et al. 1996; Payer 1999; Rambo 2001; Stokland 2001; Lohr et al. 2002; Heilmann- 22 Clausen and Christensen 2003). Forest productivity was declining in Europe (Schütt and 23 Cowling 1985) and North America (Siccama et al. 1982; Knight 1987) and the importance that 24 dead wood has on overall forest health were becoming apparent. Comparisons between managed 1 25 and natural forests of Fennoscandia (Finland, Norway, and Sweden) clearly illustrated the 26 substantial impact that forest management practices can have on wood-inhabiting organisms 27 (Andersson and Hytteborn 1991; Esseen et al. 1992; Siitonen and Martikainen 1994; Bader et al. 28 1995; Sippola et al. 1998). 29 Fennoscandian forestry historically (beginning in the 1950s) was one of the most 30 mechanized and intensive in the world (Gustafsson et al. 2010), a characteristic retained to this 31 day. Forest management in Fennoscandia often included clearcutting, site preparation, tree 32 planting, followed by multiple thinning entries throughout a rotation cycle. Almost all of the 33 forested land is used as a productive timber forest, resulting in drastically simplified forests 34 characterized by even-age, even-structure, possessing very little CWD (Esseen et al. 1992; 35 Gustafsson et al. 2010). Studies comparing saproxylic biodiversity within Finnish and Russian 36 forests (Siitonen and Martikainen 1994; Juha et al. 2002; Rouvinen et al. 2002) have noted that 37 even though both areas have similar boreal forest conditions the management practices that have 38 been applied are very different. More than 90% of the forests in Finland are productive forest, 39 with less than 3% protected park reserves (Koivula and Vanha-Majamaa 2020), while Russian 40 forests have not been harvested to the same extent due to lack of infrastructure, inefficient 41 machinery, and depressed economic conditions (Juha et al. 2002). Juha et al. (2002) found the 42 Russian forests to be more heterogenous than Finnish forests, the most significant difference 43 being the greater abundance of dead trees in Russia. Siitonen and Martikainen (1994) found 2 44 saproxylic species considered extinct in Finland and 5 considered endangered within 20 working 45 hours in Russia. These intense management practices have left 217 species red listed in Finland 46 caused by the reduction of CWD, and 317 species red listed due in part to the reduced amounts 47 of CWD (Siitonen 2001). Finland and other northern European countries have since adopted 2 48 retention forestry practices, with greater effort to conserve and even create CWD (Halme et al. 49 2013; Gustafsson et al. 2020). 50 Forests in Canada have not been managed to the same extent as most northern European 51 countries. Most logging still occurs in “primary” forests, though there is increasing harvesting in 52 second-growth forests (Berch et al. 2012). In 1995 the ecological value of CWD was recognized 53 within BC forestry legislation, however there was a “zero waste tolerance policy” until 1998 that 54 required licensees to remove all wood above certain size and soundness limits while harvesting 55 otherwise they were monetarily penalized (Caza 1993). Starting in 1999, waste benchmarks, 56 which included CWD, were established to allow a volume of waste to be left without being 57 monetarily billed, these amounts were based on economics and not based on scientific studies 58 (BC Ministry of Forests, Lands, Natural Resource Operations and Rural Development 2005). 59 The current waste benchmark ranges from 4–20 m3·ha-1 depending on the forest type (BC 60 Ministry of Forests, Lands, Natural Resource Operations and Rural Development 2005). The 61 Forest and Range Practices Act (FRPA) requires licensees to retain a minimum of 4 logs per 62 hectare, each being a minimum of 5 m in length on the Coast, or 2 m in length in the Interior, and 63 30 cm in diameter at one end on the Coast, or 7.5 cm at one end in the Interior (FPPR 2004). 64 Currently the Chief Forester’s guidance suggests retaining large (length and diameter) pieces, a 65 range of decay classes, overlapping logs that have some elevation above the ground, and dead 66 standing trees that will provide future CWD (BC Ministry of Forests, Lands, Natural Resource 67 Operations and Rural Development 2010). BC’s forest management has had a long history of 68 conflicting CWD requirements and penalizations (Stone et al. 2002), which has led to 69 inconsistent and poor CWD management. Ongoing research on the specific quantities and 70 qualities of CWD required to support ecosystem functions (Lofroth 1993; Jönsson and Jonsson 3 71 2007; Seip et al. 2018; Proulx and Aubry 2020), aims to increase our understanding of the roles 72 that CWD plays and will hopefully be integrated into policy and improve CWD management. 73 The aim of this thesis is to advance our understanding of how timber harvesting practices 74 impact CWD dynamics through time and contribute to the development of CWD landscape level 75 management tools such as remote sensing airborne laser scanning (ALS). The study area is 76 located at the Date Creek Silvicultural Systems Study, in the Kispiox Valley in northwestern BC. 77 Date Creek is a research forest dedicated to evaluating the long-term effects of retention 78 harvesting on many ecological values. My research evaluates the temporal effects of timber 79 harvesting on CWD quantity and quality with a specific focus on the habitat contributions CWD 80 provides to marten, Martes americana. This thesis comprises two data chapters written as 81 manuscripts. In chapter 1, which was published in the Canadian Journal of Forest Research 82 (Farnell et al. 2020), I evaluate the impact of retention forestry relative to clear cutting or 83 unmanaged forests on CWD attributes and the contributions CWD makes to habitat features 84 important to marten over a 27-year period. I examine four CWD components: (1) volume and 85 decay class, (2) inputs from windthrow, (3) contributions to habitat structure of the American 86 marten and, (4) CWD occurrence across stands varying in age and productivity. In chapter 2, I 87 evaluate how ALS-derived stand structure metrics can be used across the variety of historic 88 clearcut, partial cut, and natural areas to: a) determine CWD volume, b) identify CWD attributes 89 known to be important for marten habitat, and c) examine how the differences between the 90 varied previous partial cuts can be identified at the landscape scale. Finally, I synthesis chapters 91 1 and 2 in a concluding chapter. 4 92 2. Chapter 1: The effects of variable retention forestry on coarse woody debris dynamics 93 and concomitant impacts on American marten habitat after 27 years. 94 Published as: Farnell, I., Elkin, C., Lilles, E., Roberts, A.-M., and Venter, M. 2020. The effects 95 of variable retention forestry on coarse woody debris dynamics and concomitant impacts on 96 American marten habitat after 27 years. Can. J. For. Res. 50(9): 925–935. Doi:10.1139/cjfr- 97 2019-0417. 98 99 Abstract: Coarse woody debris (CWD) in the form of logs, downed wood, stumps, and large 100 tree limbs is an important structural habitat feature for many small mammal species, including 101 the American marten. At a long-term experimental trial in northern temperate hemlock-cedar 102 forests of British Columbia, Canada, we analysed the impact of varying amounts of overstory 103 basal area retention: 0% (clearcut), 40%, 70%, 100% (unharvested) on CWD volume, decay 104 class, and inputs from windthrow over 27 years. We used CWD attributes (diameter, length, 105 decay class, and height above the ground) known to be favourable for marten to create an index 106 for assessing the impact of harvesting intensity on CWD habitat features. Stands with 70% 107 retention had CWD attributes that resulted in CWD habitat features similar to unharvested 108 stands. Clearcuts contained pieces that were smaller, more decayed and closer to the ground, 109 which contributed less valuable habitat, compared to stands with higher retention. Over the 27- 110 year period, windthrown trees were the majority of CWD inputs and volume change was 111 positively related to percent retention. Our results highlight that forest management influences 112 CWD size and input dynamics over multiple decades, and the need for consideration of these 113 impacts when undertaking long-term multiple-use forestry planning. 5 114 Keywords: coarse woody debris, partial cutting, mustelids, American marten habitat, retention 115 forestry, windthrow 116 2.1. Introduction 117 Coarse woody debris (CWD) is an important component of many forest ecosystem 118 functions (Harmon et al. 1986), and there is increasing recognition that CWD dynamics need to 119 be considered during forest management planning. CWD supports many ecosystem services: it 120 promotes biodiversity, provides structural complexity (Spears et al. 2003), contributes to nutrient 121 cycling (Brunner and Kimmins 2003), influences species composition (Feller 2003), impacts soil 122 and sediment transport (Short et al. 2015), and stores carbon (Magnússon et al. 2016). CWD is 123 also a critical habitat component for many species of plants, invertebrates (Schiegg 2000), fungi, 124 algae, bacteria, bryophytes (Ódor et al. 2006), and lichen, that rely on it for suitable growing 125 sites and nutrient sources (Stokland et al. 2012). For vertebrates such as mustelids (Wiebe et al. 126 2014), small mammals (Sullivan et al. 2017), and reptiles (Sandström et al. 2019) CWD is a vital 127 habitat component as it provides resting, denning and nesting sites, access to subnivean 128 thermoregulation and prey sites, and cover from predators. Specifically, CWD is an important 129 habitat element for mustelids such as the American marten, Martes americana, a common 130 indicator species for sustainable forest management throughout much of North America 131 (Thompson et al. 2012). 132 The ability of CWD to provide these ecosystem functions depends not only on the 133 amount, but also the distribution and state (i.e., size, decay status, orientation, and position) of 134 the pieces of decaying wood, all of which are influenced by forest structure and the natural 135 disturbances and harvesting practices that initiate CWD recruitment (Brassard and Chen 2006). 136 In the temperate hemlock-cedar forests of northern British Columbia, CWD additions have 6 137 historically come from natural disturbances such as mixed or high severity fire, windstorms, bark 138 beetles, insect defoliators, and root rot (Daust and Price 2014). Fires create large amounts of 139 CWD, initially as snags, which will eventually become CWD as they fall to the forest floor 140 within 10–50 years (Foster et al. 1998). These new inputs add to the residual CWD, as most 141 previous CWD is not consumed in fires (Brown et al. 2003). CWD generated from wind storms 142 differs from fire in that the dead wood is added directly to the ground (Spies and Cline 1988). 143 Windthrow, insects, and disease often kill single or groups of trees, creating structural and spatial 144 diversity throughout the stand, though they occasionally cause catastrophic tree mortality. In 145 general, CWD produced by natural disturbances is often elevated off the ground or supported by 146 the surrounding trees. 147 In contrast, timber harvesting creates CWD primarily as fine woody debris or slash, 148 which is often small and rests on the ground (Pedlar et al. 2002), and which typically has a 149 dispersed distribution over the landscape (Siitonen et al. 2000). However, there can be 150 considerable variation in CWD inputs both within and between clearcut operations, as processing 151 can be done at the stump or roadside, the latter creating large slash piles near the road which are 152 often burned. Harvesting equipment contributes to the mechanical reduction of logging debris 153 and legacy CWD, which often results in the CWD being less elevated above the ground and 154 more fragmented (McCarthy and Bailey 1994). CWD decay rates and losses may be altered by 155 harvest dependent micro-site climate shifts, changes in substrate quality (Harmon et al. 1986), 156 and the relative elevation of CWD pieces from the ground (Kaytor 2016). Harvest operations 157 will also shift future CWD addition rates by altering both the abundance and condition of the 158 remaining trees (Densmore et al. 2004). For example, short rotation times of 100 years or less 159 reduce senescence and natural disturbances effectively eliminating some of the key processes 7 160 that contribute to CWD production (Moroni 2006). Previous studies have found that after one or 161 two harvesting rotations little of the preharvest CWD remains and further accumulation is very 162 low (Densmore et al. 2004). 163 In contrast with clearcutting, silviculture practices that incorporate some basal area 164 retention offer the potential to maintain higher CWD dependent ecosystem services, while still 165 extracting economic value from the forests (Lindenmayer et al. 2012). Partial harvesting systems 166 retain both singular individuals and groups of trees in larger quantity than clearcuts and can 167 contribute to post-harvest structural diversity (Gustafsson et al. 2012). Spatial variability of 168 CWD will depend on the pattern of partial harvesting and how it is implemented across the 169 landscape. Depending on the extent of removal, retention forestry has varying impacts on CWD 170 inputs (Fraver et al. 2002) as tree retention provides an input source of diverse CWD pieces over 171 the rotation period (Stevenson et al. 2006). However, it is still unclear what the long-term impact 172 of different harvesting practices are on CWD characteristics and specifically the contribution that 173 CWD makes to habitat quality for other species. 174 We use the American marten as a representative species for describing the contributions 175 that CWD attributes make to habitat quality for mustelids and other CWD dependent species 176 (Sullivan and Sullivan 2020). The functional importance of CWD to martens is well documented 177 and includes access to prey (Godbout and Ouellet 2010) and subnivean thermoregulation resting 178 sites (Mclaren et al. 2013), cover from predators (Hodgman et al. 1994), hunting cues and 179 hunting efficiency (Andruskiw et al. 2008), territorial scent marking sites (Porter et al. 2005), 180 and denning and resting sites (Bull and Heater 2000). Martens have been found to prefer 181 structurally sound, long (Seip et al. 2018), large diameter (>20 cm) (Wiebe et al. 2014) CWD 182 pieces that are elevated off the ground (Corn and Raphael 1992). Stands with >200 m3·ha-1 of 8 183 CWD have been identified as preferred sites in sub-boreal coniferous forests (Lofroth 1993). 184 Marten are generally considered to be reliant upon late-successional coniferous forests (Buskirk 185 and Powell 1994), however, research has demonstrated that they exhibit some plasticity of their 186 habitat use, using mixed woods and regenerating forests (Potvin et al. 2000; Poole et al. 2004). 187 For example, Porter et al. (2005) documented a marten population successfully inhabiting a 188 young deciduous forest where they preferentially used habitat structures characteristic of older 189 forests. Other studies reporting marten use of younger stands suggest that downed logs, standing 190 snags, and a high conifer content (Chapin et al. 1997; Payer and Harrison 2003; Thompson et al. 191 2008) are the key to the effective use of younger stands. However, there is no evidence that large 192 marten populations can be sustained over the long term in intensively managed forests 193 (Thompson et al. 2008). The characteristics and amount of CWD characteristics that contribute 194 to quality marten habitat, also benefit many other species and functions (Bunnell and Houde 195 2010; Sullivan and Sullivan 2020). 196 In this study we evaluate the impact of retention forestry relative to clear cutting or 197 unmanaged forests on CWD attributes and the contributions CWD makes to habitat features 198 important to marten over a 27-year period. We examine four CWD components: (1) volume and 199 decay class, (2) inputs from windthrow, (3) contributions to habitat structure of the American 200 marten and, (4) CWD occurrence across stands varying in age and productivity. 201 2.2. Methods 202 2.2.1. Study area 203 204 All data were collected within the Date Creek Silvicultural Systems Study, located in the Kispiox Valley of northwestern British Columbia (55˚ 22’ N, 127˚ 50’ W) and within Gitxsan 9 205 Laxwiipip (house territories) Xsa Gay Laaxan and Tsihl Hlii Din. This operational scale forestry 206 experiment was established in 1992 to examine alternatives to traditional clearcut harvesting 207 (Coates et al. 1997). The research area is in the interior cedar hemlock biogeoclimatic zone 208 (moist cold Hazelton variant – ICHmc2; Meidinger and Pojar 1991), which has a transitional 209 climate between the mild, wet weather of north coastal BC, and the drier, more continental 210 weather of the interior plateau. The elevation gradient spans from 359 to 669 m and the area 211 receives an average of 535 mm of annual precipitation, 184 cm of annual snowfall and 426 mm 212 of annual rainfall, with a mean annual temperature of 4.4 ˚C (Coates et al. 1997). 213 The Date Creek research area is principally comprised of mature and old growth forests 214 that originated from stand-replacing fires. The mature forests, established after fires in 1834 and 215 1855, are dominated by western hemlock (Tsuga heterophylla [Raf.] Sarg.), western red-cedar 216 (Thuja plicata Donn ex D. Don in Lamb), and hybrid spruce (a complex of Picea glauca 217 [Moench] Voss, Picea sitchensis [Bong.] Carr. and Picea engelmannii Parry ex Engelm.) with 218 six minor tree species: subalpine fir (Abies lasiocarpa [Hook.] Nutt.), amabilis fir (Abies 219 amabilis [Dougl. ex Loud.] Forbes), lodgepole pine (Pinus contorta var. latifolia Engelm.), paper 220 birch (Betula papyrifera Marsh.), trembling aspen (Populus tremuloides Michx.), and black 221 cottonwood (Populus balsamifera ssp. Trichocarpa Torr. & Gray). Within the study area old 222 growth forests, established after a fire 350–370 years ago, and are dominated by western 223 hemlock and western redcedar with minor components of amabilis and subalpine fir (Coates et 224 al. 1997). 225 2.2.2. Experimental design 226 227 Three treatments with different canopy retention levels were applied at Date Creek in 1992 and compared to an unharvested control: 0% (clearcut) retention; 40% retention; and 70% 10 228 retention (Coates et al. 1997). In the clearcut (0% retention) all conifer trees were removed, few 229 scattered aspen and birch trees were retained, and very few understory trees survived. In the 40% 230 retention treatment trees were removed in large gaps (0.05–0.5 ha) in combination with single 231 tree or small gaps (<0.03 ha) that were evenly distributed across the stand. In the 70% retention 232 treatment trees were removed in infrequent large gaps and frequent single tree or small gaps. 233 Harvesting in the clearcut and 40% retention treatments was carried out by conventional (hand 234 falling and line skidding) and mechanical methods (feller-buncher, grapple and line skidders, and 235 a delimber) and harvesting in the 70% retention treatment was performed by hand falling and 236 horse or small equipment skidding. Harvesting occurred in the summer-winter of 1992/3 and the 237 areas were replanted in spring of 1993. 238 Before harvesting, edaphic grids of relative moisture and nutrients were mapped as 239 described in Coates et al. (1997). The study area was then stratified by stand age and soil 240 moisture regime into four forest types: old growth (350 + years old) forest with moist soils 241 (mesic 350 years old), mature forest with moist soils (mesic 140 years old), mature forest with 242 moist and wetter soils (mesic-subhygric 140 years old), and mature forest with moist and drier 243 soils (mesic-submesic 140 years old). The four treatments were applied to ~20 ha stands in each 244 forest type in a randomized complete block design, for a total of 16 treatment units in the 245 experiment. 246 2.2.3. Data collection 247 CWD measurements over 27 years 248 249 In this study CWD is considered dead wood that is not self-supporting with a tilt angle <45˚ from the ground plane and with a diameter ≥10 cm, including otherwise supported or 11 250 entangled fallen trees not yet resting on the forest floor. The line intersect technique (Van 251 Wagner 1968) was used to sample CWD pre-treatment (year 0), and at 1, 19, and 27 years post- 252 treatment. Ten random transect clusters per treatment unit, with three 30 m transects each, were 253 sampled at year 0 and 1. Five stratified random clusters per treatment unit for clearcut and 254 unharvested treatments and seven clusters per treatment unit for 40% and 70% retention 255 treatments were sampled at year 19 and resampled at year 27 (clearcuts were not sampled at year 256 19 and five treatment units had only three or four clusters sampled at year 27). 257 Diameter, species, and decay class (1–4) were recorded for each piece intersecting the 258 transect at the intersection point. Transect distances were corrected for slope for years 19 and 27, 259 and volumes were corrected for tilt angle in year 19. Length was not recorded so an inclusion 260 probability adjustment was not calculated for the decay class analysis. 261 CWD piece attribute measurements 262 At year 26, detailed CWD measurements were recorded in each treatment unit using 263 fixed-area 10 m x 10 m plots. Five plots (except for 40% retention treatment units) were 264 randomly located within a 30-m buffer zone from the treatment unit boundary in a fixed 265 orientation. In the 40% retention units 10 sample plots were established; 5 in the matrix and 5 in 266 the harvested gaps to assure both the matrix (retention) and gaps were sampled. Treatment unit 267 CWD totals were then adjusted for area in gaps versus matrix forest. In addition to species, total 268 length and decay class (1–5), and diameter and height off the ground at both ends, were recorded 269 for each piece. Pieces were included if the piece lay within or intersected the plot boundary and 270 was <50% embedded in the ground. With circular plots, the inclusion zone for each piece is 12 271 sausage shaped (Gove and Van Deusen 2011), but with square plots it is shaped like the 2D 272 projection of a cuboid. 273 Windthrow 274 Live stems >10 cm dbh that were uprooted or snapped off by wind events were surveyed 275 along belt transects in the 40%, 70%, and 100% retention treatment units at year 2, 5, 12, and 21 276 (see Coates et al. 2020 for detailed methods). The volume of these stems that became CWD was 277 estimated using total inside bark volume equations for BC tree species (Nigh 2016), and bark 278 thickness and bark volume equation for BC tree species (Kozak and Yang 1981). Locally 279 calibrated diameter-height relationships were used to estimate tree height for years without field 280 measurements (years 12 and 21). For trees that were snapped off, a snap height of 5 m (the mean 281 snapped snag height in a 2018 survey) was taken off the bottom before volume was calculated. 282 Windthrow was not monitored in the clearcuts but was assumed to be zero for all sampling years 283 because there were very few stems (>10 cm dbh) that could have potentially blown over until 284 year 18 when permanent sample plots were established. Between year 18 and year 26, no trees in 285 the clearcut permanent sample plots became snags, so any inputs to CWD from windthrow 286 would have been negligible. 287 2.2.4. Data analysis 288 CWD volume calculation 289 290 Volume (m3·ha-1) of CWD in each transect in each sampling year was calculated as: [1] ( ·ℎ ) = ×∑ ( 13 ) 291 where L is the total length of the transect (horizontal distance in m), Di is the diameter of each 292 piece of CWD (cm), and Ai is the tilt angle from horizontal for each piece (degrees) (Van 293 Wagner 1968). Because tilt angle was not measured in 1992, 1993, and 2019 it was assumed to 294 be zero, so that cos(Ai) = 1 for all pieces in those years. 295 CWD habitat index 296 A CWD habitat index was created to estimate and quantify the contribution that 297 individual CWD pieces make to marten habitat based on four CWD attributes: large end 298 diameter, height above ground, total piece length, and decay state. This index combines these 299 multiple attributes into one measure that can be assessed throughout time. It does not indicate 300 overall marten habitat suitability but rather how CWD attributes within the stand contribute to 301 important habitat features. We applied the CWD habitat index to all CWD pieces in our fixed- 302 area plots measured in year 26. Individual pieces were assessed and measurements normalized so 303 that each piece was given a value between 0 and 1. The index included the following criteria 304 based on the literature: 305 (a) Coarse woody debris pieces >20 cm in diameter [where di is the diameter (d) of piece i] can 306 increase habitat suitability, as recommended by Wiebe et al. (2014) and Payer and Harrison 307 (2003). Habitat value increases with diameter, because larger pieces are more likely to be used as 308 den sites, have more subnivean space, tend to have more subnivean access points, and higher 309 marten hunting success (Andruskiw et al. 2008), and the rate of decay decreases with diameter 310 (Herrmann et al. 2015) which influences the duration of a piece of downed wood. Above a 311 diameter of 1 m all pieces are considered to have the maximum value: 14 = 1 ⁄100 0 ≥ 100 ≥ 20 < 20 < 100 312 [2] 313 (b) CWD pieces on the ground (hi = 0, where hi is height in centimeters above the ground on the 314 higher end) are not as valuable as elevated pieces which allow greater access to subnivean 315 habitats (Mclaren et al. 2013). Height above the ground (hi) was measured on the higher end of 316 each CWD piece. Habitat value increases with height above ground up to 100 cm, above which 317 all pieces have the maximum value, as snow pack in the area rarely reaches 100 cm (Wang et al. 318 2016): 1 = ℎ ⁄200 + 0.5 0.5 ℎ ≥ 100 ℎ >0 ℎ ≤0 319 [3] ℎ 320 (c) CWD pieces of decay class (ci) 1, 2, or 3 are structurally useful for martens, (Spencer et al. 321 1983), but their value decreases with increasing decay state as they are less structurally sound 322 and will be retained for less time (Lofroth 1993). As such, the contribution each piece makes 323 decreases linearly with increasing decay from a maximum of 1 for class 1 to a minimum of 0.5 324 for class 3: 325 [4] 326 (d) Longer CWD pieces are used as runways and refuges in winter by marten (Bunnell and 327 Houde 2010). Large pieces contribute to piles of stacked CWD, may be used as den structures 328 and support structural complexity preferred by marten (Bull and Heater 2000). Pieces >2 m (BC 329 minimum CWD length regulations) are useful but longer pieces have higher value as Seip et al. = 1.25 − 0.25 ∗ 0 ≤3 >3 15 330 (2018) found the probability of marten presence increased exponentially with increasing mean 331 CWD length: 332 [5] 333 where li is the length of pieces, and l95th is the 95th percentile length of all pieces measured (in our 334 data this was 23m). 335 A per hectare CWD habitat index value, CWD-HI, was calculated for each plot by multiplying 336 the ratio of the averaged normalized piece metrics to the piece’s inclusion zone, ai, and 337 multiplying by 10,000. 338 [6] 339 where s is the side length of the square plots (10 m) and i is the length of each piece, 340 [7] 341 Statistical analysis ⁄ = - = 0 ≤2 <2 ⁄ = 10,000 ∑ + 342 We examined the effect of harvesting retention level and time on CWD volume and 343 decay class using linear mixed effects models with the “lmerTest” package v. 3.1 (Kuznetsova et 344 al. 2017) in R version 3.6.1 (R Core Team 2020) using the Kenward-Roger degrees-of-freedom 345 method. The fixed effects were treatment, year, and the treatment × year interaction. Forest type 346 was included as a random factor, with year nested within treatment, and treatment nested within 347 forest type. We did not account for model residual error with a special repeated measures 348 structure, but a separate analysis indicated that the residual first order auto-correlation was very 349 low across year 0, 1, or 27. Marginal means were estimated for within year treatment effects 16 350 using the R package “emmeans” (Lenth 2019); a Tukey familywise error adjustment corrected p- 351 values for multiple comparisons of treatments within years. 352 To test for treatment effects on the CWD habitat index and CWD piece attributes (decay 353 class, diameter, height above the ground, length, number of pieces, and number of pieces >20 cm 354 in diameter) in year 26 we used a model with treatment specified as a fixed effect which was 355 nested within the forest type random effect. Unlike the above model, which included the 356 observations of plots within treatment units, plots in year 26 were averaged within treatment 357 units (in order to weight the 40% retention treatment unit means by the area in the harvested gap 358 versus matrix forest). Mean estimations and multiple comparisons were done using the same 359 methods as above. We further analysed the effect of forest type on CWD volume, decay class, 360 and habitat index, by extracting the best linear unbiased predictors (BLUPs) from the overall 361 model (this was done using the R package “nlme”; Pinheiro et al. 2016). 362 2.3. Results 363 2.3.1. Harvesting effects on CWD volume through time 364 CWD volume among the harvest treatments diverged over 27 years, with higher volumes 365 accompanying higher retention (Figure 1), although the treatment × time interaction was only 366 marginally significant (Table 1). Before harvest and immediately post-harvest CWD volume did 367 not differ among treatments (Figure 1). However, by year 27, volume was twice as high in the 368 unharvested stands as the clearcuts and the 40% and 70% volumes were intermediate (Figure 1). 369 The average annual CWD change after harvest was -1.1 m3·ha-1·year-1 in the clearcuts, -0.2 370 m3·ha-1·year-1 in the 40%, 1.5 m 3·ha-1·year-1 in the 70%, and 2.4 m3·ha-1·year-1 in the unharvested 371 stands (Appendix 1, Table 1). 17 372 Table 1: Results of linear mixed effects models (F stat, numerator (ndf) and denominator (ddf) 373 degrees of freedom and p-value). Measurements are missing for the 0% retention stands in year 374 19. Response variable Volume (m3 ha-1) Year(s) postharvest 0, 1, 19, 27 Decay class 0, 1, 19, 27 Habitat quality Decay class Total length Height above ground Diameter # of pieces > 10 cm # of pieces > 20 cm 26 26 26 26 26 26 26 Factor Treatment Year Treatment × Year Treatment Year Treatment × Year Treatment Treatment Treatment Treatment Treatment Treatment Treatment 375 376 18 F (ndf,ddf) 4.5 (3,9.7) 8.3 (3,35.6) 2.0 (8,34.1) 0.8 (3,9.4) 89.5 (3,33.6) 2.6 (8,33.2) 19.6 (3,9) 4.3 (3,9) 13.4 (3,9) 9.8 (3,9) 2.0 (3,9) 0.2 (3,9) 6.1 (3,9) p-value 0.03 0.0003 0.08 0.5 <0.0001 0.03 0.0003 0.04 0.001 0.003 0.2 0.9 0.01 377 378 Figure 1: Coarse woody debris volume (m3·ha-1) in each treatment pre-harvest (0), 1, 19- and 379 27-years post-harvest (A) and cumulative windthrow volume (m3·ha-1) in the 40%, 70% 380 retention, and unharvested stands 2, 5, 12, and 21 years post-harvest. Volumes are presented as 381 boxplots of the treatment units with the medians, quartiles and outliers. The lower-case letters 382 represent the within year treatment effect’s significance, treatments with the same letter are not 383 different. 384 2.3.2. Harvesting effects on CWD decay class through time 385 The average decay class of CWD pieces depended on the interaction between harvest 386 treatment and time over 27 years (Table 1). Immediately after harvest the clearcut and 40% 387 retention stands had a higher proportion of pieces in decay class 1 than the unharvested stands 388 which had pieces more evenly distributed among decay classes (Figure 2). Before harvest, 19 19 389 years, and 27 years post-harvest a large proportion of the pieces were in decay class 3 across 390 treatments. At year 27 we could not detect a difference in decay class among treatments with the 391 transect method, however, the fixed plot method did indicate that decay class varied with 392 retention level at year 26 (Table 1). By this time the clearcuts had a smaller proportion of less 393 decayed pieces than the unharvested stands (Table 2). 394 Table 2: Piece characteristics means and ± standard errors in year 26. The lower-case letters 395 represent the within year treatment effect significance; treatments with the same letter are not 396 different. Treatment 0% 40% 70% Unharvested Length (m) 4.08 ± 0.43 a 6.07 ± 0.43 b 6.04 ± 0.43 b 7.95 ± 0.43 b Diameter (cm) 19 ± 0.014 a 18 ± 0.014 a 21 ± 0.014 a 20 ± 0.014 a Decay class (1–5) 3.5 ± 0.1 a 3.2 ± 0.1 ab 3.2 ± 0.1 ab 3.1 ± 0.1 b 397 20 Height above ground (m) 0.07 ± 0.031 a 0.17 ± 0.031 ab 0.20 ± 0.031 b 0.28 ± 0.031 b # of pieces > 10 cm 719 ± 82 a 639 ± 82 a 700 ± 82 a 724 ± 82 a # of pieces > 20 cm 212 ± 25 ab 159 ± 25 a 298 ± 25 b 207 ± 25 ab 398 399 Figure 2: Proportion of coarse woody debris (CWD) pieces in each decay class in each 400 treatment pre-harvest (0), 1, 19, and 27 years post-harvest. Decay classes are presented in a 401 violin plot, which show the sample’s distribution density using the width of the plot, and the 402 median. The width of the plot also presents the number of pieces in each treatment year; skinnier 403 plots have less pieces of CWD compared to wider plots. Lower-case letters represent the 404 treatment effect’s significance, treatments with the same letter are not significantly different; 405 years without letters indicate an insignificant treatment effect. 406 2.3.3. Windthrown CWD inputs through time 407 Across the treatments windthrow accounted for most of the CWD inputs over time, but 408 recruitment varied by treatment. In the unharvested stands, windthrow of 1.0 m3·ha-1·year-1 was 21 409 less than half of the total volume gain (Appendix 1, Table 1 & 2). In the 70% retention stands, 410 windthrow of 1.1 m3·ha-1·year-1 accounted for majority of total volume gain (Appendix 1, Table 411 1 & 2). In the 40% retention stands, windthrow inputs of 1.1 m 3·ha-1·year-1 were not enough to 412 balance the loss from decay and total volume decreased over time (Appendix 1, Table 1 & 2). 413 The absence of CWD inputs from windthrow in the clearcuts contributed to the total volume loss 414 of 1.1 m3·ha-1·year-1 (Appendix 1, Table 1 & 2). 415 2.3.4. Harvesting effects on CWD attributes, and the marten CWD habitat index, 26 years 416 post-harvest 417 Twenty-six years post-harvest, the CWD habitat index differed among treatments (Table 418 1). The index was less than half as high in the clearcuts compared to the 70% retention and 419 unharvested stands (Figure 3). The contribution of CWD pieces to habitat quality was lower in 420 the clearcuts because CWD pieces were more decayed, shorter, and closer to the ground (Table 421 2; Figure 4). Although the 40% retention stands had longer pieces than the clearcut stands (Table 422 2) we could not detect a difference in CWD habitat contribution between the two treatments 423 (Figure 3). Compared to the unharvested, CWD pieces in the 70% retention stands contributed as 424 much or more to marten habitat (Figure 3), with pieces of similar length, diameter, decay, and 425 height above ground (Table 2). The 70% retention and unharvested stands were the only 426 treatments with pieces elevated over 4 m (Figure 4). The CWD piece diameters on average were 427 not different among treatments nor were the total number of pieces >10 cm, but the number of 428 pieces >20 cm diameter, which contributed to the CWD habitat index, were highest in the 70% 429 retention stands and lowest in the 40% retention stands (Table 2). 22 430 431 Figure 3: Habitat quality in year 26 post-harvest; presented as boxplots of the individual plots 432 within treatments (plots in the 40% retention are a combination of the matrix and harvested gaps) 433 with the medians, quartiles and outliers; lower-case letters represent the treatment effect 434 significance; treatments with the same letter are not significantly different. A single outlier of 435 300 was recorded in the 70% treatment that is not visible with the axis scaling. 23 436 437 Figure 4: Distribution of coarse woody debris (CWD) piece characteristics: A) decay class (1– 438 5); B) total length (m); C) height above the ground (m); and D) diameter (cm) that contribute to 439 the habitat quality index. Characteristics of individual CWD pieces within treatments are 440 presented violin plots, which show the sample’s distribution density using the width of the plot 441 and the median. 442 2.3.5. Variation in CWD across forest types 443 Forest type as a random factor accounted for approximately 25% of the observed 444 variation in CWD volume (variance represented by forest type was 2723.1, while residual 445 variance was 7743.9). Point estimates of the forest type random factor, using best linear 446 unbiased prediction, indicated that forest type had a significant effect on CWD volume (p = 24 447 <0.001, F3,9 = 46.1). Volumes were highest in the mesic 350 year block, intermediate in the 448 mesic-subhygric 140 year block, and lowest in the mesic and mesic-submesic 140 year blocks 449 (Figure 5). The old growth (mesic 350 year) stands typically had over twice the volume of CWD 450 as the mesic and submesic 140 year blocks (Figure 5; Appendix 1, Table 1). Although CWD 451 volume varied between forest types, the treatment effect was consistent across forest types: the 452 random effect of treatment nested within forest type accounted for only 4% of the observed 453 variation in CWD volume. 454 Unlike volume, less than 1% of the variation in decay class was accounted for by forest 455 type. Although CWD volume differed among forest types, the average decay class of those 456 pieces was the same. Forest type accounted for a large amount of the observed variation in the 457 habitat index (approximately 46%) but forest type was only marginally significant according to 458 BLUPs (p = 0.07, F3,9 = 3.3). Similar to CWD volume, the habitat index tended to be higher in 459 the mesic 350 year block followed by the mesic-subhygric 140 year block. Old growth and 460 higher productivity stands tended to have CWD with better characteristics for marten. 25 461 462 Figure 5: Coarse woody debris volume (m3·ha-1) in each forest type pre-harvest (0), 1, 19, and 463 27 years post-harvest. Volumes are presented as boxplots of the treatment units within blocks 464 with the medians, quartiles, and outliers. 465 2.4. Discussion 466 Results from our study indicate that retaining overstory basal area in a partial-cutting 467 system facilitates the development of CWD characteristics favorable to mustelids such as 468 marten. Immediately following harvest there were no differences in CWD volume between the 469 retention treatments. The CWD benefits of retention were accrued through time and increases in 470 CWD volume were observed in stands with higher retention 27 years post-harvest. In addition, 471 retention level impacted the structural attributes of the CWD and by extension the value of its 472 contribution to marten habitat over time. We found that harvest impacts on CWD are not simply 26 473 driven by short term (immediately following harvest to 5 years post-harvest) differences in CWD 474 input rates, but also by longer term shifts in the stand structure dependent processes that 475 influence CWD recruitment, such as windthrow. 476 Many studies, across different forest types and stand ages, observe a pulse of CWD 477 immediately after harvest that results in an increase in volume (McCarthy and Bailey 1994; 478 Feller 2003; Densmore et al. 2004; Moroni 2006) due to waste from harvest operations, though 479 this is not always consistent (Fridman and Walheim 2000; Stevenson et al. 2006). We did not 480 detect a significant volume increase after harvest, however there was an increase in the 481 proportion of pieces in decay class 1. In accord with Stevenson et al. (2006) and Fraver et al. 482 (2002) we found the proportion of class 1 pieces increased with higher harvesting levels. After 483 harvest most new class 1 pieces came from windthrow, which meant there were very few class 1 484 pieces recruited to the clearcuts over time, and average decay class was higher in clearcuts than 485 unharvested stands at year 26. The 40% and 70% retention stands continued to recruit CWD and 486 resumed their pre-harvest decay class distribution. We anticipate the CWD volume in the 487 clearcuts to continue to diverge from the stands with retained structure as time progresses, due to 488 the residual pieces continuing to decay and fewer inputs being created. We also expect the 489 effects of partial harvesting to become more evident on CWD volume as time progresses, but this 490 process may be slow. Over a longer time period than this study (46 years), Morrissey et al. 491 (2014) observed a significantly lower CWD volume in partial-cut compared to unharvested 492 stands. Partial-cutting altered species composition and structure which decreased windthrow and 493 senescence-related tree mortality (Morrissey et al. 2014). We have not yet detected similar 494 decreases in mortality at Date Creek, but these dynamics will continue to be monitored. 27 495 A wide range in CWD volume has been reported for temperate hemlock-cedar forests 496 (81.67–387 m3·ha-1; Densmore et al. 2004; Stevenson et al. 2006), and we also found a wide 497 range in volume in unharvested stands (57.7–326.1 m3·ha-1). Forest type accounted for much of 498 this variability, suggesting that CWD is not evenly distributed across the landscape. Feller (2003) 499 found that CWD volume in British Columbian old-growth forests increased by an average of 15 500 m3·ha-1 for each 1 m increase in forest productivity. In our study older and more productive 501 stands had higher CWD volumes. In contrast, studies in other forest types, such as deciduous old 502 growth, no variation in CWD volume was observed across different forest communities and 503 topographic positions (Davis et al. 2015). Regardless, high CWD volumes in older and more 504 productive sites is associated with good habitat for marten and other CWD dependant organisms 505 as suggested from our results and other studies (Ódor et al. 2006; Stevenson et al. 2006; Ylisirniö 506 et al. 2009). Stands are not equal in CWD quantity, CWD attributes, or CWD spatial distribution 507 which suggests that spatially explicit CWD management should consider productivity as well as 508 stand age. However, we found the relative effect of harvesting intensity on CWD, was consistent 509 across the forest types we studied. 510 While total CWD volume is important for understanding some ecosystem functions such 511 as carbon storage (Magnússon et al. 2016) and nutrient dynamics (Laiho and Prescott 2004), 512 CWD piece attributes are important for understanding forest ecosystem services related to stand 513 structure (Sandström et al. 2019). The CWD habitat index we created is a simple way to assess 514 the attributes of CWD pieces that contribute to structural complexity that supports habitat 515 functions. This index could be used by forest managers to assess baseline CWD conditions, 516 develop management targets or to assess CWD before and after management activities. While 517 the index was parameterized for temperate hemlock-cedar forests, the framework could be 28 518 adjusted for other forest types or specific interests. For example, in areas such as some boreal 519 forests where CWD rarely exceeds 20 cm in diameter (Greif and Archibold 2000), the minimum 520 diameter could be adjusted to suit local marten requirements. To include marten prey (mice and 521 voles) habitat in the index all decay classes could have value since small mammals utilize CWD 522 of all decay classes (Fauteux et al. 2012; Sullivan and Sullivan 2020). Fauteux et al. (2012) 523 found that some small mammal species were positively associated with logs of early decay 524 classes. Although early decay class logs can provide cover against predation, well-decayed CWD 525 may offer more fungi and invertebrates for food, higher humidity and more nesting opportunities 526 for small mammals. 527 Many studies have shown that marten prefer habitat with a complex physical structure, 528 especially near the forest floor (Payer and Harrison 2003; Godbout and Ouellet 2010). We found 529 26 years post-harvest, CWD in the clearcut contributed less to the habitat index because the 530 majority of pieces were short, decayed, and resting on the ground. In contrast to clearcuts, we 531 found that the 40% and 70% retention and unharvested treatments showed continual recruitment 532 of CWD. These new recruits mostly came from previously retained trees uprooting or snapping 533 mid-bole (Coates et al. 2020). Freshly windthrown trees are long, structurally sound and are 534 more likely to be elevated off the ground. These pieces can become entangled and stacked, 535 which creates complex vertical structures resulting in a higher contribution to marten habitat 536 (Payer and Harrison 2000). 537 Several recent studies (Andruskiw et al. 2008; Sullivan and Sullivan 2019, 2020) have 538 examined the response of small mammal communities to forest harvesting practices, specifically 539 comparing marten use of retention patches with their use of clearcuts in which constructed 540 woody debris piles were developed. While the use of forest patches in a harvest system 29 541 maintained small mammals and mustelids, including marten, the abundance and community of 542 small mammals in clearcuts varied in treatments with constructed CWD structures compared to 543 dispersed CWD arrangements (Sullivan and Sullivan 2019, 2020). Species diversity and 544 abundance was significantly higher in clearcuts with constructed CWD structures, however the 545 areas with dispersed wood still maintained small mammal populations. Overall, structural habitat 546 complexity enhanced the efficiency of predatory search by martens. In regenerating forests, 547 subnivean access points created by CWD are essential for small mammals, and by extension 548 important for marten. Andruskiw et al. (2008) showed that marten may cue in on CWD 549 structures that protrude above snow packs for hunting access; clearcuts that only have low-lying 550 CWD would not provide this cue for marten. 551 Our study shows that CWD that contributes to marten habitat was less after a quarter 552 century in stands managed with a clearcut system. However, stands managed with a high 553 retention or partial cutting system maintained CWD with attributes which contributed to higher 554 quality marten habitat. Acknowledging that the lifespan of CWD is finite, as it is slowly 555 reclaimed through biological and mechanical deterioration, is critical in recognizing the 556 importance of continuous CWD inputs in forest ecosystems. Our study focused on an ecological 557 function specific to one species, but many of the other functions CWD supplies will also decline 558 in harvested stands that are not managed to maintain recruitment over the whole forest rotation. 559 30 560 3. Chapter 2: Area-based approach for modeling coarse woody debris characteristics using 561 LiDAR-derived measures of forest structure 562 Abstract 563 Coarse woody debris (CWD) is an important component of forest ecosystems which 564 provides critical habitat features for many mammals, such as the American marten, Martes 565 americana. Timber harvesting impacts the amount and quality of the pieces of CWD, though this 566 impact can be reduced by using partial cutting harvesting practices. To facilitate landscape level 567 planning that aims to maintain habitat, it is vital to be able to assess where areas of high habitat 568 quality are. In this study we use airborne laser scanning (ALS) to detect differences in CWD 569 volume, habitat index value, piece characteristics (diameter, length, height above the ground, and 570 decay class) in stands of varying levels of basal area retention (0%, 40%, 70%, and 100%) at the 571 Date Creek research area. We used a machine learned (ML) model to predict the different CWD 572 attributes over the landscape and identify “hotspots” of high CWD habitat value for marten. The 573 model was able to differentiate the treatments comparable to the field data suggesting this is a 574 useful tool for landscape level planning. 575 Key words: coarse woody debris, landscape level planning, airborne laser scanning, 576 habitat, marten 577 3.1. Introduction 578 Advanced forest age and high structural diversity enhance biodiversity by increasing the 579 quantity and variety of microclimate niches (air humidity and temperature) and light conditions 580 that contribute to species richness and the overall functioning of forest ecosystems (Lindenmayer 581 and Franklin 2002; McGee and Kimmerer 2002; Taboada et al. 2010). Dead fallen trees, known 31 582 as coarse woody debris (CWD), are one component of structural complexity that increases with 583 age. CWD is ecologically important for nutrient cycling (Minnich et al. 2020), carbon storage 584 (Magnússon et al. 2016), and providing habitat for bryophytes (Ódor et al. 2006), lichen 585 (Spribille et al. 2008), fungi (Abrego et al. 2017), invertebrates (Lassauce et al. 2011), dead- 586 wood decaying microorganisms (Benbow et al. 2020), birds (Lohr et al. 2002), and many 587 mammals (Keisker 2000). 588 While CWD volume is an important metric, piece characteristics such as diameter, 589 length, height above the ground, and decay class are just as important. These CWD 590 characteristics determine if and how CWD positively influences habitat value for many small 591 mammals and mustelids, such as American marten (Martes americana) (Keisker 2000). Large 592 diameter pieces (>20 cm) remain on the landscape longer (Herrmann et al. 2015) and are more 593 likely to be used as dens, increase subnivean access points, and increase hunting success rates 594 (Andruskiw et al. 2008). Longer pieces (>2 m) are used as runways and refuges by marten in the 595 winter (Bunnell and Houde 2010) and they are more likely to be stacked, supporting preferred 596 structural complexity (Bull and Heater 2000). Pieces elevated off the ground decay slower 597 (Kaytor 2016), retaining wood strength and provide greater structural complexity which in turn 598 creates subnivean access points (Mclaren et al. 2013). These CWD characteristics can be 599 incorporated together to form a CWD habitat suitability index that can be evaluated at a plot, and 600 potentially landscape level, to assess the habitat value of CWD for marten and other CWD- 601 dependent mammals such as, fishers, mice, and voles (Farnell et al. 2020). 602 Forest management practices directly impact the total amount and quality of individual 603 pieces of CWD in a forest stand (Pedlar et al. 2002). In British Columbia (BC), Canada, from 604 1998 – 2017 the percent of timber harvested on public lands used the following allocations: 3.4% 32 605 retention system, 4.4% other forms of partial cutting (seed tree, shelterwood, selection, thinning), 606 62% clearcut with reserves and, 30% clearcut without reserves (Province of BC 2019, Beese et 607 al. 2019). The most common practices, clearcut with and without reserves, removes all the 608 standing trees (except those remaining in reserves), indiscriminately gathering damaged boles 609 and debris to be burnt; these practices remove live and dead standing trees which would 610 otherwise restock the CWD supply, and remove existing legacy pieces of CWD. Historically in 611 BC there have been minimum CWD levels required to be left post-harvest by law (FPPR 2004), 612 however the quality of these pieces is often low, lacking elevation, longevity, or habitat value 613 (Pedlar et al. 2002). There is increasing pressure from the public and supporting research to 614 increase the prevalence of less destructive timber harvesting practices such as variable retention 615 logging/partial harvesting (Bauhus et al. 2009; Király et al. 2013) and the creation of CWD 616 corridors within clearcuts (Seip et al. 2018). 617 Natural stand dynamics create varying amounts of CWD over time. Typically, 618 disturbance agents such as windstorms, insects, and root and bole disease kill trees in clumps that 619 creates stand-level variation of CWD (Vanderwel et al. 2013; White et al. 2015). However, large 620 insect outbreaks or wildfires can kill trees across the landscape, eventually creating large 621 quantities of CWD (Bassett et al. 2015; Stevens-Rumann et al. 2015). Generally, CWD follows a 622 “U”-shaped pattern with forest succession after a stand replacing disturbance (Harmon et al. 623 1986; Warren, G. R. and Meades, J. P. 1986; Spies et al. 1988; Sturtevant et al. 1997; Brassard 624 and Chen 2008). From time since the event, the amount of CWD is dependant on the initial stand 625 disturbance. This residual CWD declines over time, as the stand matures small-scale 626 disturbances occur and larger quantities of CWD accumulate, with a decrease sometimes 627 reported in very old stands (Harmon et al. 1986; Spies and Cline 1988; Spies et al. 1988; 33 628 Sturtevant et al. 1997; Brais et al. 2005; Brassard and Chen 2006). However, this “U”-shaped 629 pattern is not always observed (Hély et al. 2000). Other factors, like edaphic conditions and 630 slope, also influence CWD dynamics. More CWD can be found at the bottom of a steep slope 631 (Rubino and McCarthy 2003) and on more productive sites (Spies et al. 1988; Robertson and 632 Bowser 1999; McCarthy et al. 2001) 633 Partial harvesting retains live and dead trees, removing only select valued trees for 634 harvest, which in turn retains CWD recruits and post harvest structure (Gustafsson et al. 2012). 635 The abundance of live and dead tree habitat elements is proportional to the percent retention 636 (Huggard et al. 2009; Farnell et al. 2020). Partial harvesting has been reported to maintain 637 marten habitat if basal area is >18 m2·ha-1 and canopy closure is >30% (Payer and Harrison 638 2003, 2005; Fuller and Harrison 2005), because CWD is retained and overstory canopy closure 639 and vertical structure is maintained. Along with variable retention logging there are other 640 methods of retaining CWD after harvest such as creating piles of dead wood throughout the 641 clearcut to connect areas of retained trees. Seip et al. (2018) created CWD corridors containing 642 30–50 m3·ha-1 of CWD post clearcut, higher than the current ~5 m3·ha-1 retained by industrial 643 operations. They found these corridors enhanced habitat within clearcuts for marten and their 644 prey. 645 In the late 1990s variable retention harvesting became more widely practiced by 646 commercial forest companies in response to numerous negative impacts resulting from large- 647 scale clearcutting (Beese et al. 2019). As a result, research areas such as the Date Creek 648 Silvicultural Systems Research Forest, were established throughout BC to examine the effects of 649 retention harvesting on biodiversity and forest regeneration (Coates et al. 1997). Foresters and 650 industry are now reviewing these partially cut forests for further harvest of timber, which 34 651 requires the identification of areas of rich biodiversity and good stand structure to be retained by 652 informed harvest practices in order to maintain the intended function of the retained trees. 653 Tools such as airborne laser scanning (ALS) are becoming increasingly popular to 654 understand and review spatially explicit information to assist in identifying areas of high 655 ecological value. Field surveys are often limited in their spatial extent due to site accessibility, 656 time, and cost. Supplementing field surveys with ALS has proven to be increasingly fruitful 657 (White et al. 2013). Currently forest planners use ALS to derive digital elevation models for 658 block layout and road location planning; canopy height models for tree heights, block boundary 659 locations and visual impact assessments; and hillshade rasters to extract slope data (White et al. 660 2013). Over the last two decades ALS has successfully been used to measure CWD volume 661 through individual piece detection (Blanchard et al. 2011; Mücke et al. 2013; Lindberg et al. 662 2013; Nyström et al. 2014; Joyce et al. 2019; Jarron 2020) from ALS-derived forest metrics and 663 an area-based regression to predict plot-level CWD volume (Pesonen et al. 2008; Sumnall et al. 664 2016) and multispectral LiDAR (Queiroz et al. 2020). 665 In a previous study at the Date Creek Research Forest (Farnell et al. 2020), we assessed 666 the effects of varied timber harvesting retention amounts on CWD volume, CWD habitat 667 suitability for marten, and piece characteristics (diameter, length, decay, and height above the 668 ground) over 27 years at the plot scale. In this study we explore whether ALS can detect the 669 same effects of varied timber retention and effectively assist in wildlife focused landscape level 670 planning. Here we evaluate how ALS-derived stand structure metrics can be used across a 671 variety of historic clearcut, partial cut, and natural areas to: a) determine CWD volume; b) 672 identify CWD attributes known to be important for marten habitat; and c) examine how the 35 673 differences between the varied previous partial cuts can be identified at the landscape scale and 674 be used to inform landscape level habitat planning. 675 3.2. Methods 676 3.2.1. Study area 677 All data was collected within the Date Creek Silvicultural Systems Study, located in the 678 Kispiox Valley in northwestern British Columbia (55˚ 22’ N, 127˚ 50’ W) and within Gitxsan 679 Laxwiipip (house territories) Xsa Gay Laaxan and Tsihl Hlii Din. The research area is in the 680 interior cedar hemlock biogeoclimatic zone (moist cold Hazelton variant – ICHmc2; Meidinger 681 and Pojar 1991), which has a transitional climate between the mild, wet weather of north coastal 682 BC, and the drier, more continental weather of the interior plateau. The elevation gradient spans 683 from 359 m to 669 m and the area receives an average of 535 mm of annual precipitation, 184 684 cm of annual snowfall and 426 mm of annual rainfall, with a mean annual temperature of 4.4 ˚C 685 (Coates et al. 1997). 686 The Date Creek research area is principally comprised of mature and old growth forests. 687 The mature forests are dominated by western hemlock (Tsuga heterophylla [Raf.] Sarg.), western 688 red-cedar (Thuja plicata Donn ex D. Don in Lamb), and hybrid spruce (a complex of Picea 689 glauca [Moench] Voss, Picea sitchensis [Bong.] Carr. and Picea engelmannii Parry ex Engelm.) 690 with six minor tree species: subalpine fir (Abies lasiocarpa [Hook.] Nutt.), amabilis fir (Abies 691 amabilis [Dougl. ex Loud.] Forbes), lodgepole pine (Pinus contorta var. latifolia Engelm.), paper 692 birch (Betula papyrifera Marsh.), trembling aspen (Populus tremuloides Michx.), and black 693 cottonwood (Populus balsamifera ssp. Trichocarpa Torr. & Gray). The old growth forests are 36 694 dominated by western hemlock and western redcedar with minor components of amabilis and 695 subalpine fir (Coates et al. 1997). 696 3.2.2. Experimental design 697 Three treatments with different canopy retention levels were applied at Date Creek in 698 1992 and compared to an unharvested control: 0% (clearcut) retention; 40% retention; and 70% 699 retention (Coates et al. 1997). Before harvest the study area was stratified by stand age and soil 700 moisture regime into four forest types: old growth (350 + years old) forest with moist soils 701 (mesic 350 years old), mature forest with moist soils (mesic 140 years old), mature forest with 702 moist and wetter soils (mesic-subhygric 140 years old) and mature forest with moist and drier 703 soils (mesic-submesic 140 years old). The four treatments were applied to ~20 ha stands in each 704 forest type in a randomized complete block design, for a total of 16 treatment units in the 705 experiment. 706 3.2.3. Data acquisition 707 CWD empirical data 708 Detailed CWD measurements were recorded in each treatment unit using fixed-area 10 x 709 10 m plots to ground truth the LiDAR data. Five plots were randomly located in each of the 710 treatment units within a 30-m buffer zone from the treatment unit boundary in a fixed orientation 711 (except for within 40% retention treatment units). In the 40% retention units 10 sample plots 712 were established; 5 in the matrix and 5 in the harvested gaps to assure both the matrix (retention) 713 and gaps were sampled. Treatment unit CWD totals were then adjusted for area in gaps versus 714 matrix forest. CWD pieces with a tilt angle <45˚ from the ground plane and with a diameter ≥10 715 cm were measured. Species, total length, decay class (1–5), diameter, and height off the ground 37 716 at both ends, were recorded for each piece. Pieces were included if the piece lay within or 717 intersected the plot boundary and was <50% embedded in the ground. An inclusion probability 718 adjustment was calculated. With circular plots, the inclusion zone for each piece is sausage 719 shaped (Gove and Van Deusen 2011), but with square plots it is shaped like a two-dimensional 720 projection of a cuboid. This inclusion probability was applied to each piece of CWD for volume 721 and piece characteristic calculations. 722 ALS data acquisition and preprocessing 723 Georeferenced LiDAR point cloud data for Date Creek research area were collected in 724 August 2018 during leaf on conditions with an average point density of 19.4 points·m-2 using a 725 Riegl Q1560 dual-channel LiDAR system and a Riegle LMS-Q780 LiDAR system (Appendix 2; 726 Table 1). The raw LiDAR data was tiled, filtered for noise returns, ground classified, and height 727 normalized following standard point cloud processing routines outlined by White et al. (2013) 728 using LAStools software (Isenburg 2018). 729 A suite of 43 ALS-derived metrics based on the forest structure were calculated for each 730 plot (Appendix 2; Table 2). ALS metrics were also calculated for the entire study area at a spatial 731 grid resolution of 10 m, which corresponds to the dimensions of the field plots. These metrics 732 quantify stand structure metrics such as the height of the canopy, canopy density, canopy gaps, 733 vertical complexity from the ground to the maximum height (described in Appendix 2; Table 2). 734 All metrics, except the height filtered first single return normalized counts (described below), are 735 standard metrics that can be extracted using the lascanopy function in LAStools (Isenburg 2018). 736 Following similar methods to Jarron (2020), the height filtered first single return 737 normalized counts were calculated by first setting upper and lower height thresholds. Previous 38 738 studies which have detected individual CWD pieces have demonstrated that a height filtered 739 point cloud within the height range of CWD occurrence is important to identify and detect CWD 740 effectively (Pesonen et al. 2008; Abalharth 2013; Lindberg et al. 2013; Jarron 2020). The lower 741 threshold was set to 0.2 m, to exclude terrain points due to the uncertainty of defining the forest 742 terrain (Lindberg et al. 2013). The upper threshold was set to 1 m as >90% of the pieces were 743 below this height. We also included two other height bins from 1–2 m and 2–3 m to capture piles 744 of stacked CWD. Once the height bins were set, the point cloud was again filtered to only 745 include first single returns, as CWD is likely to return only single pulses because it is large and 746 dense compared to shrubs, which send multiple returns from partial hits off the leaves, branches 747 and twigs (Wing et al. 2012). The height filtered first single return point counts are normalized 748 by dividing it and adding it to the ground counts. 749 3.2.4. Linking empirical and ALS data 750 Estimation of CWD empirical-derived response variables 751 CWD volume, diameter, length, decay class, height above the ground, and a CWD 752 habitat index value (Farnell et al. 2020) were calculated for each of the 100 ground plots. CWD 753 volume (m3) was calculated using Fraver et al. (2007) conic-parabloid volume equation: = 754 12 (5 4 +5 + 5 4 4 ∗5 4 ) 755 where L represents the piece length (m), Db is the diameter (m) at the base and Du is the diameter 756 (m) at the upper end. 757 758 The per hectare CWD habitat index value was calculated using the criteria and equations from Farnell et al. (2020): 39 - 759 +ℎ = 10,000 + + ⁄4 760 Where dnorm is normalized diameter, hnorm is the normalized height above the ground, cnorm is the 761 normalized decay class, lnorm is the normalized length and a is the inclusion probability. See 762 Appendix 1 for details on calculations for each of these components. 763 Per hectare volume (m3·ha-1) and piece characteristics, y, were calculated by: = 764 / 1/ × 10,000 765 Where yi is the response variable (either the individual piece volume or piece characteristics) and 766 ai is the individual piece inclusion probability. 767 Estimation of CWD response variables from ALS data 768 Machine learning (ML), using Ranger (Wright and Ziegler 2017) a fast implementation 769 of Random Forest (bagging regression trees), was used to develop regression trees to model the 770 empirically measured CWD variables using ALS-derived predictor variables within the 771 “SuperLearner” package (Polley et al. 2019) in R (R Core Team 2020). A Ranger model was 772 generated for each of the CWD variables (volume, length, diameter, height above the ground, 773 decay class, and CWD habitat index). The same set of ALS-derived metrics (Appendix 1; Table 774 2) were used as the predictor variables to predict each of the response variables. Each Ranger 775 model generated 500 decision trees. Roads were excluded from the ALS data using a 10 m buffer 776 on each side. 777 778 Variable importance was assessed using the base “RandomForest” package (Liaw and Wiener 2002) in R. 40 779 Model validation 780 Model fit and predictive performance were evaluated using a K-fold cross-validation 781 framework. Within this framework the data was split into 10 random groups, each group was 782 held out of the dataset while the remaining groups were used to train the model, the model was 783 then evaluated on the predictions of the held-out data. 784 3.2.5. Landscape predictions 785 Empirically-derived treatment effects 786 Using the same analysis as Farnell et al. (2020), we examined the effect of harvesting 787 retention level on empirical CWD volume, habitat index value, and piece attributes (diameter, 788 length, height above the ground, and decay class) using linear mixed effects models with the 789 “lmerTest” package v. 3.1 (Kuznetsova et al. 2017) in R using the Kenward-Roger degrees-of- 790 freedom method. The fixed effect was treatment which was nested within the forest type random 791 effect. Marginal means were estimated treatment effects using the R package “emmeans” (Lenth 792 2019); a Tukey familywise error adjustment corrected p-values for multiple comparisons of 793 treatments. 794 ALS-derived treatment effects 795 The trained ML model was used to predict each of the different response variables for the 796 Date Creek research area. The predicted raster was clipped to each treatment unit and the median 797 was extracted. We then used the same linear mixed effects model as above to examine the effect 798 of harvesting retention level on the ALS predicted volume, habitat index value, and piece 799 attributes (diameter, length, height above the ground, and decay class). The R “raster” package 800 (Hijmans 2020) was used to generate CWD raster maps for each response variable (volume, 41 801 habitat index value, length, diameter, height above the ground and decay class) for the Date 802 Creek research area. 803 3.3. Results 804 3.3.1. ALS CWD detection 805 Predicted and observed values from the volume and habitat index models are shown in 806 Figure 6. The R2 and MSE results from the model and 10-fold cross-validation for each of the 807 response variables are shown in Table 3. The model predicted each of the response variables 808 well, however the model validation resulted in a large decrease of the R2 (Table 3). The top 5 809 important predictor variables for volume were: max, vc5, p99, vc1, and vc2 (See Appendix 2, 810 Table 3 for the other response variables). 811 812 Figure 6: Regression of predicted versus observed coarse woody debris volume (m3·ha-1) (left) 813 and habitat index value (ha-1) (right). To improve readability each graph has a point cut off. (A) 814 actual = 1337.9 m3·ha-1 predicted = 762.4 m3·ha-1, and (B) actual = 295 ha-1 predicted = 177 ha-1. 42 815 816 Table 3: Model and validated R2 and MSE results for each variable assessed. Variable Model R 2 Model MSE Validated R2 Validated MSE Volume (m3·ha-1) 0.76 6222.60 -0.079 27579.46 Habitat index value (ha-1) 0.82 322.31 0.047 1664.52 Diameter (m) 0.80 0.001 -0.068 0.003 Length (m) 0.82 1.23 0.059 6.61 Decay class (1-5) 0.81 0.049 -0.134 0.384 Height above the ground (m) 0.81 0.007 -0.034 0.039 817 818 819 3.3.2. Landscape-level ALS treatment effect detection The landscape level ALS analysis allowed us to differentiate CWD volume to an 820 equivalent degree as what could be achieved using empirical data (Figure 7). CWD volume 821 linearly increased as percent retention increased. There is a significant difference between 822 clearcuts and the 70% retention and unharvested treatments. The ALS was able to differentiate 823 the 70% retention from the clearcut while the empirical data showed no difference. 824 The habitat index value also increased linearly as percent retention increased (Figure 7). 825 The ALS was able to differentiate the 70% retention from the clearcut and the unharvested from 826 the 40% retention. The ALS model slightly over predicted both the volume and habitat value and 827 showed less variation than the empirical data. The ALS CWD models were used to develop a 828 landscape raster map of the habitat index which differentiates areas of high habitat value in dark 43 829 orange and low habitat value in light orange (Figure 8) and corresponds with the historical partial 830 harvest treatments. 831 832 833 Figure 7: (A) Boxplot of coarse woody debris empirical volume (m3·ha-1) (left) and ALS 834 predicted volume (m3·ha-1) (right). (B) Boxplot of habitat index value (left) and ALS predicted 835 habitat index value (right). The lower-case letters represent significant treatment effects; 836 treatments with the same letter are not statistically different. 44 837 838 Figure 8: Prediction of coarse woody debris habitat index values for the Date Creek research 839 area at a 10 m resolution (100 m 2 grain). Habitat values increase from light to dark orange. Each 840 treatment unit is labeled with the percent retention. 841 When the aggregated habitat index is broken down into the individual CWD attributes the 842 ALS data was able to distinguish all piece characteristic treatment effects (Figure 9). The ALS 843 detects: a diameter difference between the clearcut and the other treatments, while the empirical 844 includes this trend but does not separate the classes; a difference of piece length between the 845 clearcut, 40% and 70% retention treatments while the empirical data shows this trend but does 846 not differentiate between the clearcut and the other treatments; the same decay class difference 847 between the clearcut and unharvested as the empirical data; and a difference of height above the 45 848 ground between the clearcut, 40% and 70% retention while the empirical data only differentiates 849 the clearcut and unharvested treatments. In all cases the ALS data has less within treatment 850 variation. 851 852 Figure 9: Boxplots of empirical (left) and predicted (right) coarse woody debris piece 853 characteristics: (A) diameter (m); (B) piece length (m); (C) decay class (1-5); (D) height above 854 the ground (m). The lower-case letters represent significant treatment effects; treatments with the 855 same letter are no statistically different. 46 856 3.4. Discussion 857 We evaluated how ALS-derived stand structure metrics can be used across a variety of 858 historic clearcut, partial cut, and natural areas to predict CWD volume, habitat index value and 859 piece characteristics. We found that ALS can distinguish CWD in varied retention treatments 860 with the level of detail and results comparable to the empirical data over the landscape. The 861 ALS-derived stand structure model performs well when all the training data is used but 862 decreased when training data were withheld for validation. We believe the lower R2 values 863 associated with the validation procedures are partly due to the spatially variable nature of CWD. 864 Some areas, such as windthrow piles, have very high amounts of CWD, while 10 m next to these 865 piles could have none (Davis et al. 2015). When plots with either a high amount or no CWD are 866 left out of the training dataset the model poorly predicts these plots. 867 Landscape-level projections of CWD volume, CWD habitat index and piece 868 characteristics over the full research area demonstrated that the ALS models differentiated 869 landscape level CWD differences between different partial-harvest treatments, which 870 corresponded with the trends observed in empirical data. While landscape level estimates were 871 good, our ALS model struggled with predicting CWD volume at a fine grain (100 m2) . At a 872 larger grain size, we expect the model accuracy would converge with landscape level accuracy. 873 The challenge of developing precise ALS models that are accurate at a fine spatial grain 874 has been previously identified (Pesonen et al. 2008; Beland et al. 2019). However, ALS can be a 875 valuable tool that allows partially harvested areas to be distinguished. Using ALS to make 876 landscape level predictions provides invaluable information to forest managers that need to 877 identify what is currently on the landscape, especially for wildlife such as marten that have 878 highly variable home ranges areas (~200 – 2,000 ha) (Buskirk and McDonald 1989; Proulx and 47 879 Aubry 2020), often linked to habitat composition (Smith and Schaefer 2002). For example, 880 within the home range of martens there is a maximum proportion of the range that can be non- 881 forested or very low habitat value, if the threshold is exceeded these areas won’t be used 882 (Steventon 2014). ALS can be used to assess the amount of forested land and current habitat 883 quality to aid the decision process of where corridors of habitat should be retained when making 884 land use plans that encompass multiple values (Proulx and Aubry 2020). 885 Other remote sensing CWD methods, such as individual piece detection, have been 886 successful at estimating plot-level CWD volume (Lindberg et al. 2013; Joyce et al. 2019; Jarron 887 2020). Recently Jarron (2020) extrapolated the individual piece detection to stand-level scales 888 using a mixture of height and pulse-based filters and linear pattern recognition to vectorize 889 pieces of CWD >30 cm in diameter to measure their volume and expand to plot (R2 = 0.81) and 890 stand level volume estimates. We used similar height and pulse-based filters for three of our 891 predictor variables (norm_c0.2-1, norm_c1-2 m, norm_c2–3 m), but they did not rank high in the 892 variable importance assessment. However, the point density height filtered data without the 893 single-return filter (d0.2-1) was a strong predictor variable for habitat index, diameter, and decay 894 class. This suggests the ALS density of points between 0.2–1 m from both single and multiple 895 returns to be better predictor of CWD than single returns alone. This finding may reflect the fact 896 that within our study area, larger and older CWD were often colonized by bryophytes and/or 897 vegetation, with the consequence that multiple returns were reflected from CWD structure. Our 898 study area is in the interior cedar hemlock biogeoclimatic zone, which are the wettest and most 899 productive forests in interior BC, producing dense bryophyte cover. Individual piece detection 900 involves many processing steps, reducing its operational effectiveness through additional 901 complexity and costs. In contrast, the stand-structure metrics we used are commonly used to 48 902 assess stand variables like canopy density, basal area, canopy height, etc. Therefore, using the 903 same metrics may increase the operational ease for landscape level planning and potential for 904 integration into current planning processes. 905 Pesonen et al. (2008) used a similar area-based stand metric regression approach to 906 predict CWD. They found canopy gaps, represented by greater numbers of first pulses that 907 reflect from close to the ground, and the standard deviation in height pulses, were the most 908 powerful predictor variables of CWD volume. In our study “cov”, which represents the canopy 909 cover/gaps, was not one of the top predictor variables. This may be due to the difference in study 910 location/forest type. Pesonen et al (2008)’s study area was in a conservation area, where the gaps 911 that were created were likely from natural disturbance events where the dead trees remain at the 912 site. In our study area, which is composed of different harvesting treatments, many of the gaps 913 that were created were from timber extraction, so there are no remaining dead trees in the gap 914 location. However, 26 years has passed since time of harvest, providing time for natural tree 915 succession and CWD to be created. In our study, we used predictor variables to represent the 916 vertical complexity from the ground to the maximum height of the canopy in varying height 917 stratifications (1, 2, 5, and 10 m). Vertical complexity ranked high in the variable importance 918 assessment for volume, habitat index, length, height above the ground, and decay class. This 919 suggests that the level of homo-or-heterogeneity of the stands best predicts CWD in previously 920 harvested areas. 921 CWD characteristics other than volume should be considered when managing forests. 922 When we evaluated CWD piece characteristic, we were able to detect treatment differences 923 reasonably well using stand-structure ALS metrics, sometimes exceeding the estimates that could 924 be developed from a limited number of empirical field plots. The empirical data did not allow 49 925 diameter differences to be separated between treatments, however the ALS distinguished the 926 clearcut to have smaller diameter pieces than the other treatments. The ALS also differentiated 927 CWD diameter, length, and height above the ground in the 40% retention from the 70% retention 928 and unharvested treatments whereas sample number limitations associated with the empirical 929 data meant that variables such as CWD length and height above the ground were not 930 differentiated between the treatments using empirical data. ALS was also able to identify 931 differences in decay class between different treatments, an advantage which the area-based 932 approach presents over individual piece detection methods; advanced levels of decay reduce the 933 probability for detection (Mücke et al. 2013). As the number of bryophytes and level of decay 934 increase, the pieces become less distinguishable from the forest floor. The ability to detect 935 diameter, length, height above the ground, and decay class helps forest managers develop a 936 clearer ecological perspective of the current landscape to assist with operational planning. 937 Farnell et al. (2020) combined CWD piece attributes to create an index that quantifies the 938 contribution that individual CWD pieces make to marten and other CWD-dependant species’ 939 habitat. The functional importance of CWD to marten includes access to prey (Godbout and 940 Ouellet 2010), subnivean thermoregulation resting sites (Mclaren et al. 2013), cover from 941 predators (Hodgman et al. 1994), hunting cues (Andruskiw et al. 2008), territorial scent marking 942 sites (Porter et al. 2005), and denning and resting sites (Bull and Heater 2000). Combining the 943 piece attributes into a habitat index and using ALS can help identify the “hot-spots” of where 944 good CWD habitat features are located within the landscape by creating a visual raster map. The 945 model overestimated the habitat value of plots that had zero value, presenting a limitation. This 946 is perhaps due to the use of stand structure metrics and the habitat index in conjunction, 947 compounding the indications of poor habitat value. Plots with zero habitat value likely have 50 948 similar stand structure metrics to low habitat value plots. They may still have CWD within them, 949 but the pieces may be too small in diameter to meet the criteria to be included in the index. 950 Therefore, the model is unable to distinguish zero value habitat from low value habitat areas. 951 Other indexes could be expanded to landscape scales using an area-based approach, such as the 952 structural diversity index developed by Storch et al. (2018), which uses 11 aspects of structural 953 diversity to create an index to quantify the level of structural diversity in large-scale forest 954 inventories. Recently, de Assis Barros and Elkin (2021) created an old-growth index that 955 identifies areas of “very-high” old growth values and used a similar Random Forest modeling 956 framework to identify these values over a 18,000 ha study area. Alternatively, an index could be 957 created by mapping the individual components of the function first and then creating a 958 presence/absence criteria for mapping the final habitat suitability. Martinuzzi et al. (2009) used 959 such an approach to identify habitat suitability of snags and understory vegetation for songbirds 960 in the mixed temperate forests of northern Idaho. Landscape level indices are effective tools to 961 aid forest managers to identify and prioritize timber harvesting but also to facilitate the 962 conservation of ecologically rich sites. 963 Globally, land managers are trying to develop better ways for managing forest resources, 964 particularly in a way that will maintain multiple objectives for future forest landscapes. In BC, it 965 has only recently been highlighted that although there is a mandate to manage forests for a range 966 of different objectives, including wildlife habitat and timber, there is a significant gap between 967 forest resource planning and practice due to inconsistent landscape-level planning (BC Forest 968 Practices Board 2019). Landscape-level planning, if done effectively, translates the strategic 969 objectives, such as species at risk planning, cumulative effects, and policy development, to what 970 is feasible for harvest practices on the ground. Operational planning, such as timber harvesting 51 971 layout, can be undertaken with a holistic understanding of harvest practices on the desired 972 values. Using tools like the ALS framework we have developed will close this gap between 973 resource planning and practice by providing land managers spatially explicit data to implement 974 the multi-value objectives. Our study has shown that ALS can identify CWD attributes and areas 975 of high value CWD habitat over the landscape. Landscape level planning and future-focused 976 plans are critical in achieving desired future forests for all values (Lindenmayer et al. 2008; BC 977 Forest Practices Board 2019). 978 52 979 4. Chapter 3: Conclusions 980 4.1. Overview 981 The impact of retention harvesting on CWD volume and important piece characteristics 982 for habitat, and the ability of ALS to detect the impacts, were investigated in this thesis. The 983 effects of harvesting on CWD volume have been well studied (Edmonds, Robert L. et al. 1986; 984 Spies and Cline 1988; Hansen et al. 1991; Caza 1993; McCarthy and Bailey 1994; Sippola et al. 985 1998, 1998; Duvall and Grigal 1999; Siitonen et al. 2000; Fridman and Walheim 2000; Siitonen 986 2001; Stokland 2001; Rouvinen et al. 2002; Pedlar et al. 2002; Fraver et al. 2002; Stevenson et 987 al. 2006; Moroni 2006; Jönsson and Jonsson 2007; Brassard and Chen 2008; Vanderwel et al. 988 2008; Morrissey et al. 2014; Keren and Diaci 2018; Thorn et al. 2020; Koivula and Vanha- 989 Majamaa 2020; Wang et al. 2021) but there have been few studies on the CWD piece 990 characteristics that create valuable habitat, and on methods of assessing these characteristics over 991 the landscape. In Chapter 1, I analysed the impact of varying amounts of overstory basal area 992 retention: 0% (clearcut), 40%, 70%, 100% (unharvested) on CWD volume, decay class, and 993 inputs from windthrow over 27 years. I used CWD attributes (diameter, length, decay class, and 994 height above the ground) known to be favourable for marten to create an index for assessing the 995 impact of harvesting intensity on CWD habitat features. In chapter 2, I assess whether ALS can 996 be used to detect differences in CWD volume, habitat index value, and piece characteristics 997 (diameter, length, decay class, and height above the ground) in the stands of varying levels of 998 basal area retention. A predictive map of CWD habitat value was created across the Date Creek 999 landscape that highlights areas of high CWD habitat value. 1000 4.2. Key findings 53 1001 Stands with 70% retention had CWD attributes that resulted in CWD habitat features 1002 similar to unharvested stands. Clearcuts contained pieces that were smaller, more decayed, and 1003 closer to the ground, which contributed less valuable habitat, compared to stands with higher 1004 retention. Over the 27-year period, windthrown trees contributed the majority of CWD inputs, 1005 and volume change was positively related to percent retention. Using an ALS area-based 1006 approach poorly predicts CWD at the plot (100 m2) level (Chapter 2: Table 3), however it 1007 successfully predicts CWD volume, piece characteristics, and habitat value at the landscape 1008 level. The ALS predictive model was able to predict similar CWD volume, habitat value, and 1009 piece characteristic trends in each of the retention treatments to the empirical data. The “5 m 1010 vertical complexity” predictor variable was the top predictor variables for 4 out of 6 response 1011 variables, suggesting that stand homo-or-heterogeneity is strongly related to CWD volume and 1012 piece characteristics in previously harvested areas. 1013 4.3. Implications 1014 Forest managers who aim to manage for timber as well as other biodiversity objectives 1015 need to consider how much basal area is being removed and what the implications will be. The 1016 results from Chapter 1 indicate that retaining overstory basal area (40% and 70%) during timber 1017 harvesting facilitates the development of CWD characteristics favourable to CWD-dependant 1018 species such as marten. It is important to maintain high volumes of CWD as well as high quality 1019 pieces - large diameter, long, elevated pieces within the full range of decay classes. Twenty-six 1020 years after harvest, these attributes did not develop in stands that were clearcut, while the partial 1021 retention stands maintained these attributes at levels comparable with unharvested stands. The 1022 habitat index I created is a useful way to combine each of these attributes into one value that can 54 1023 be used to assess baseline CWD conditions, develop management targets, or to assess CWD 1024 before and after management activities. 1025 Chapter 2 results indicate that ALS is a powerful tool that can be used to detect CWD 1026 volume and the quality of the pieces over the landscape. However, as the methods I used for 1027 predicting CWD are poor at predicting at the 100 m2 plot level, the scale at which ALS is used 1028 should be considered when using these methods. The importance for landscape level planning, 1029 especially in areas that have previously been harvested, is increasing, as well as more pressure to 1030 manage for multiple values. The methods developed here demonstrate the ability of ALS to 1031 assess CWD volume and it’s piece characteristics, at a stands scale (i.e. Date Creek research 1032 forest), and provide valuable landscape management information. 1033 4.4. Limitations and future research 1034 The lifespan of CWD is very long, often exceeding 100 years. The 27 years of 1035 monitoring CWD at Date Creek is one of the longest CWD monitoring studies, however 27 years 1036 is not long-term in the lifespan of CWD. CWD should continue to be monitored at Date Creek to 1037 study the full impact of retention harvesting, especially if another partial harvest is to occur in 1038 the 40% and 70% retention stands. 1039 The CWD marten habitat index only considers the CWD habitat component for marten 1040 and disregards other important aspects of their habitat such as overhead canopy cover and prey 1041 abundance. Future research could include this habitat index into other existing marten models, or 1042 other habitat components could be integrated into this index. The model is also specific to 1043 marten, which could be adapted to better suite other CWD dependant species such as fishers or 1044 beetles by adjusting the parameters for each of the piece characteristics. 55 1045 The ALS model will preform best at the landscape scale and when used for similar forest 1046 types. For example, caution should be used if the model was applied to dissimilar forest or 1047 forests that have not undergone a retention harvest. Increasing the number of plots and the plot 1048 area to 200 m2 may increase the ability of ALS to estimate smaller scales. Sampling a greater 1049 variety of forest types, e.g., recently clearcut stands, would increase the forest variation the ALS 1050 can accurately predict. 1051 There has been almost a century of research on the topic of CWD, with studies 1052 concluding that CWD is essential for biodiversity and overall forest health, and negative impacts 1053 timber harvesting has on CWD and its associated biodiversity. The current policies surrounding 1054 CWD retention do not reflect its importance. This must change for species like marten to remain 1055 in healthy populations across their full habitat range. “Dead, damaged, or diseased, it stays” 1056 (Franklin 2019). 1057 56 Appendix 1 Table 1: Coarse woody debris volume (m3·ha-1, mean and ± standard error) (diameter >10 cm) by treatment unit, preharvest (1992), one-year post-harvest (1993), 19 years post-harvest (2011) and 27 years post-harvest (2019). Treatment Forest type 0% retention Mesic 140 yr Mesic 350 yr Mesic-subhygric 140 yr Mesic-submesic 140 yr 1992 57.7 ± 23.6 120.2 ± 18.6 133.1 ± 37.8 80.3 ± 12.4 Volume (m3·ha-1) 1993 2011 71.5 ± 13.6 182.8 ± 25.7 144.7 ± 16.7 94.2 ± 11.2 Mesic 140 yr 40% retention Mesic 350 yr Mesic-subhygric 140 yr Mesic-submesic 140 yr 89.5 ± 17.0 163.3 ± 41.1 116.0 ± 24.5 50.8 ± 6.8 124.0 ± 30.5 289.2 ± 55.2 119.2 ± 25.5 83.7 ± 11.3 106.6 ± 14.8 273.4 ± 82.9 139.4 ± 23.7 76.4 ± 17.0 122.5 ± 21.2 257.4 ± 84.4 130.8 ± 28.7 83.6 ± 19.3 Mesic 140 yr 70% retention Mesic 350 yr Mesic-subhygric 140 yr Mesic-submesic 140 yr 61.7 ± 11.1 132.9 ± 18.7 84.1 ± 11.9 94.3 ± 24.2 79.7 ± 18.3 229.3 ± 48.9 110.9 ± 18.8 101.2 ± 12.8 136.6 ± 57.8 201.7 ± 24.8 170.4 ± 29.7 164.0 ± 14.8 150.4 ± 65.7 193.5 ± 27.1 164.2 ± 28.7 167.7 ± 19.0 Unharvested 101.6 ± 23.1 285.5 ± 44.6 156.8 ± 35.6 87.2 ± 19.8 111.3 ± 26.8 189.6 ± 23.3 149.2 ± 20.3 130.0 ± 33.1 211.4 ± 27.6 329.9 ± 56.9 237.1 ± 49.3 120.3 ± 24.5 158.4 ± 25.0 326.1 ± 67.0 235.0 ± 38.1 110.1 ± 22.5 Mesic 140 yr Mesic 350 yr Mesic-subhygric 140 yr Mesic-submesic 140 yr 57 2019 39.1 ± 12.9 155.0 ± 21.3 128.7 ± 31.3 58.1 ± 11.4 Table 2: Cumulative windthrow volume (m3·ha-1) by treatment unit, two years post-harvest (1994), five years post-harvest (1997), 12 years post-harvest (2004), and 21 years post-harvest (2013). Cumulative volume (m3·ha-1) 1994 1997 2004 2013 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Treatment 0% retention Forest type Mesic 140 yr Mesic 350 yr Mesic-subhygric 140 yr Mesic-submesic 140yr 40% retention Mesic 140 yr Mesic 350 yr Mesic-subhygric 140 yr Mesic-submesic 140yr 4.2 13.4 0.9 2.0 4.8 15.0 10.7 3.2 16.1 17.8 17.3 4.2 25.6 34.5 26.5 7.0 70% retention Mesic 140 yr Mesic 350 yr Mesic-subhygric 140 yr Mesic-submesic 140yr 2.6 20.5 5.5 6.7 2.9 22.6 7.2 8.2 4.7 31.6 16.1 17.8 7.3 31.6 28.7 27.3 Unharvested Mesic 140 yr Mesic 350 yr Mesic-subhygric 140 yr Mesic-submesic 140yr 2.7 13.3 5.6 1.3 4.0 14.6 7.9 1.7 10.6 16.6 18.0 6.3 13.1 27.0 33.9 11.2 58 Appendix 2 Table 1: ALS acquisition characteristics for Date Creek research area. Characteristic Sensor Wavelength Flying altitude Flying speed Scan rate Scan angle Minimum overlap Average point density Average pulse density 2018 LiDAR Riegl Q1560 and Riegle LMS-Q780 1064 nm 1450 m 140 kts nominal 800 khz (533 khz usable) 58˚ 72% 19.4 points·m-2 12 m-2 Table 2: ALS metrics extracted using the lascanopy function from LAStools. ALS metric Abv All Min Max Avg Qav Std Kur Hom p01 p05 p10 p25 p50 p75 p90 p99 b10 b20 Definition The total number of points that are above the height cut off of 1.37 m. The total number of points including those that are below the cut off. The minimum height above the height cut off of 1.37 m. The maximum height. The average of all heights above the height cut off of 1.37 m. The average square height. The standard deviation of all heights above the height cut off of 1.37 m. The kurtosis. “Height of Median Energy”. All points above the height cutoff are ordered by their elevation. Then the height is computed at which the sum of intensities of points below and the sum of intensities of points above is identical. Height 1 st percentile. Height 5 th percentile. Height 10 th percentile. Height 25 th percentile. Height 50 th percentile. Height 75 th percentile. Height 90 th percentile. Height 99 th percentile. The percentage of points between 0 m and 10% of the maximum height. The percentage of points between 0 m and 20% of the maximum height. 59 b30 b40 b50 b60 b70 b80 b90 Cov Dns VC1 VC2 VC5 VC10 c_ground c0.2-1 c1-2 c2-3 d0.2-1 d1-2 d2-3 Norm_c0.21 Norm_c1-2 Norm_c2-3 The percentage of points between 0 m and 30% of the maximum height. The percentage of points between 0 m and 40% of the maximum height. The percentage of points between 0 m and 50% of the maximum height. The percentage of points between 0 m and 60% of the maximum height. The percentage of points between 0 m and 70% of the maximum height. The percentage of points between 0 m and 80% of the maximum height. The percentage of points between 0 m and 90% of the maximum height. Canopy cover – computed as the number of first returns above the height cut off divided by the number of all first returns and output as a percentage. Canopy density – computed as the number of all points above the height cut off divided by the number of all returns. Vertical Complexity Index using 1 m height stratifications. A number between 0 and 1. If VCI is close to 1 the height bins have similar point densities (homogenous). If VCI is close to 0 the distribution of the height bins is more uneven (heterogenous). Vertical Complexity Index using 2 m height stratifications. Vertical Complexity Index using 5 m height stratifications. Vertical Complexity Index using 10 m height stratifications. The number of ground points. The number of first and single return points between 0.2–1 m. The number of first and single return points between 1–2 m. The number of first and single return points between 2–3 m. The number of first return points between 0.2–1 m divided by the total number of points and scaled to a percentage. The number of first return points between 1–2 m divided by the total number of points and scaled to a percentage The number of first return points between 2–3 m divided by the total number of points and scaled to a percentage Normalized point count between 0.2–1 m. c0.2-1/(c0.2-1 + c_ground) Normalized point count between 1 –2 m. c1-2/(c1-2 + c_ground) Normalized point count between 2 –3 m. c2-3/(c2-3 + c_ground) 60 Table 3: Top 5 predictor variables for each of the coarse woody debris response variables from the machine learned model. Response Variable Volume (m3·ha-1) CWD habitat index (ha-1) Diameter (m) Length (m) Height above the ground (m) Decay class (1–5) Top 5 predictor variables max, vc5, p99, vc1, vc2 vc5, vc10, max, d0.2-1, p99 all, std, max, d0.2-1, cov vc5, b90, std, kur, vc10 vc5, b90, std, kur, vc10 c_ground, vc10, d02.-1, max, std 61 References Abalharth, M.H. 2013. Using LiDAR to detect in-stream woods: a scaled approach. Doctoral dissertation, University of British Columbia, Vancouver, BC. Abrego, N., Christensen, M., Bässler, C., Ainsworth, A.M., and Heilmann-Clausen, J. 2017. Understanding the distribution of wood-inhabiting fungi in European beech reserves from species-specific habitat models. Fungal Ecology 27: 168–174. doi:10.1016/j.funeco.2016.07.006. Andersson, L.I., and Hytteborn, H. 1991. Bryophytes and decaying wood ─ a comparison between managed and natural forest. Ecography 14(2): 121–130. doi:10.1111/j.16000587.1991.tb00642.x. Andruskiw, M., Fryxell, J.M., Thompson, I.D., and Baker, J.A. 2008. Habitat-mediated variation in predation risk by the American marten. Ecology 89(8): 2273–2280. doi:10.1890/07-1428.1. Arthur, M.A., and Fahey, T.J. 1990. Mass and nutrient content of decaying boles in an Engelmann spruce - subalpine fir forest, Rocky Mountain National Park, Colorado. Canadian Journal of Forest Research 20(6): 730–737. doi:doi.org/10.1139/x90-096. de Assis Barros, L., and Elkin, C. 2021. An index for tracking old-growth value in disturbance-prone forest landscapes. Ecological Indicators 121: 107175. doi:10.1016/j.ecolind.2020.107175. Bader, P., Jansson, S., and Jonsson, B.G. 1995. Wood-inhabiting fungi and substratum decline in selectively logged boreal spruce forests. Biological Conservation 72(3): 355–362. doi:10.1016/0006-3207(94)00029-P. Bassett, M., Chia, E.K., Leonard, S.W.J., Nimmo, D.G., Holland, G.J., Ritchie, E.G., Clarke, M.F., and Bennett, A.F. 2015. The effects of topographic variation and the fire regime on coarse woody debris: Insights from a large wildfire. Forest Ecology and Management 340: 126–134. doi:10.1016/j.foreco.2014.12.028. Bauhus, J., Puettmann, K., and Messier, C. 2009. Silviculture for old-growth attributes. Forest Ecology and Management 258(4): 525–537. doi:10.1016/j.foreco.2009.01.053. BC Forest Practices Board. 2019. Tactical forest planning: the missing link between strategic planning and operational planning in BC. Special Report. Available from https://www.bcfpb.ca/wpcontent/uploads/2019/07/SR58-Tactical-Forest-Planning.pdf. BC Ministry of Forests, Lands, Natural Resource Operations and Rural Development. 2005. Provincial logging residue and waste management procedures manual. Timber Pricing Branch. Available from https://www2.gov.bc.ca/assets/gov/farming-natural-resources-andindustry/forestry/timber-pricing/residue-and-waste/rwp_amend_28.pdf. BC Ministry of Forests, Lands, Natural Resource Operations and Rural Development. 2010. Coarse woody debris. Available from https://www.for.gov.bc.ca/ftp/hfp/external/!publish/frep/extension/Chief%20Forester%20shor t%20CWD.pdf. Beese, W.J., Deal, J., Dunsworth, B.G., Mitchell, S.J., and Philpott, T.J. 2019. Two decades of variable retention in British Columbia: a review of its implementation and effectiveness for biodiversity conservation. Ecological Processes 8(1): 33. doi:10.1186/s13717-019-0181-9. Beland, M., Parker, G., Sparrow, B., Harding, D., Chasmer, L., Phinn, S., Antonarakis, A., and Strahler, A. 2019. On promoting the use of lidar systems in forest ecosystem research. Forest Ecology and Management 450: 117484. doi:10.1016/j.foreco.2019.117484. Benbow, M.E., Barton, P.S., Ulyshen, M.D., Beasley, J.C., DeVault, T.L., Strickland, M.S., Tomberlin, J.K., Jordan, H.R., and Pechal, J.L. 2020. Necrobiome framework for bridging decomposition ecology of autotrophically and heterotrophically derived organic matter. Ecology. doi:10.1002/ecm.1331. 62 Berch, S.M., Curran, M., Dymond, C., Hannam, K., Murray, M., Tedder, S., Titus, B., and Todd, M. 2012. Criteria and guidance considerations for sustainable tree stump harvesting in British Columbia. Scandinavian Journal of Forest Research 27(8): 709–723. doi:10.1080/02827581.2012.725768. Blanchard, S.D., Jakubowski, M.K., and Kelly, M. 2011. Object-based image analysis of downed logs in disturbed forested landscapes using lidar. Remote Sensing 3(11): 2420–2439. doi:10.3390/rs3112420. Bowman, J.C., Sleep, D., Forbes, G.J., and Edwards, M. 2000. The association of small mammals with coarse woody debris at log and stand scales. Forest Ecology and Management 129(1–3): 119– 124. doi:10.1016/S0378-1127(99)00152-8. Brais, S., Sadi, F., Bergeron, Y., and Grenier, Y. 2005. Coarse woody debris dynamics in a post-fire jack pine chronosequence and its relation with site productivity. Forest Ecology and Management 220: 216–226. doi:10.1016/j.foreco.2005.08.013. Brassard, B.W., and Chen, H.Y.H. 2006. Stand structural dynamics of North American boreal forests. Critical Reviews in Plant Sciences 25(2): 115–137. doi:10.1080/07352680500348857. Brassard, B.W., and Chen, H.Y.H. 2008. Effects of forest type and disturbance on diversity of coarse woody debris in boreal forest. Ecosystems 11(7): 1078–1090. doi:10.1007/s10021-008-9180-x. Brown, J.K., Reinhardt, E.D., and Kramer, K.A. 2003. Coarse woody debris: managing benefits and fire hazard in the recovering forest. General Technical Report RMRS-GTR-105, USDA Rocky Mountain Research Station, Ogden, UT. Brunner, A., and Kimmins, J.P. 2003. Nitrogen fixation in coarse woody debris of Thuja plicata and Tsuga heterophylla forests on northern Vancouver Island. Canadian Journal of Forest Research 33: 1670–1682. doi:10.1139/x03-085. Bull, E.L., and Heater, T.W. 2000. Resting and denning sites of American marten in northeastern Oregon. Northwest Science 74(3): 179–185. Available from http://hdl.handle.net/2376/1018 Bull, E.L., Parks, C.G., and Torgersen, T.R. 1997. Tree and logs important to wildlife in the interior Columbia River basin. General Technical Report. PNW-GTR-391, United States Department of Agriculture Forest Service, Portland, OR. Bunnell, F.L., and Houde, I. 2010. Down wood and biodiversity — implications to forest practices. Environmental Reviews 18(NA): 397–421. doi:10.1139/A10-019. Buskirk, S.W., and McDonald, L.L. 1989. Analysis of variability in home-range size of the American marten. The Journal of Wildlife Management 53(4): 997. doi:10.2307/3809601. Buskirk, S.W., and Powell, R.A. 1994. Habitat ecology of fishers and American martens. In Martens, Sables, and Fishers: Biology and Conservation. Edited by S.W. Buskirk, A.S. Harestad, M.G. Raphael, and R.A. Powell. Cornell University Press, Ithaca, NY. pp. 283–296. Buskirk, S.W., and Ruggiero, L. 1994. American Marten. In The scientific basis for conserving large carnivores: American marten, fisher, lynx, and wolverine in the western United States. Edited by L. Ruggiero, K. Aubrey, S. Buskirk, J. Lyon, and W. Zielinski. United States Department of Agriculture. pp. 17–37. Caza, C.L. 1993. Woody debris in the forests of British Columbia: a review of the literature and current research. Ministry of Forests Research Program, Victoria, B.C. Available from https://www.for.gov.bc.ca/hfd/pubs/docs/mr/Lmr078.htm. Chapin, T.G., Harrison, D.J., and Phillips, D.M. 1997. Seasonal habitat selection by marten in an untrapped forest preserve. The Journal of Wildlife Management 61(3): 707–717. doi:10.2307/3802178. Coates, K.D., Banner, J.A., Steventon, D., LePage, P., and Bartemucci, P. 1997. The Date Creek Silvicultural Systems Study in the interior cedar-hemlock forests of northwestern British Columbia: overview and treatment summaries. BC Ministry of Forest Research Program. 63 Available from http://www.for.gov.bc.ca/hfd/pubs/docs/lmh/Lmh38.pdf [accessed 18 May 2018]. Coates, K.D., Lilles, E.B., Dhar, A., and Hall, E.C. 2020. Wind damage over 21 years across different levels of tree removal in natural-origin mixed forests of northwestern British Columbia. Canadian Journal of Forest Research 50(9): 946–952. doi:dx.doi.org/10.1139/cjfr-2019-0359. Corn, J.G., and Raphael, M.G. 1992. Habitat characteristics at marten subnivean access sites. The Journal of Wildlife Management 56(3): 442–448. doi:10.2307/3808856. Daust, D., and Price, K. 2014. Natural disturbance. In A climate change vulnerability assessment for British Columbia’s managed forests. Available from https://www2.gov.bc.ca/gov/content/environment/natural-resource-stewardship/naturalresources-climate-change/natural-resources-climate-change-applied-science. Davis, J.G., Chapman, J.I., Wu, S.-Y., and McEwan, R.W. 2015. Spatiotemporal dynamics of coarse woody debris in an old-growth temperate deciduous forest. Forest Science 61(4): 680–688. doi:10.5849/forsci.14-156. Densmore, N., Parminter, J., and Stevens, V. 2004. Coarse woody debris inventory and decay modelling. BC Journal of Ecosystems and Management 5(2): 14–29. Available from https://jemonline.org/index.php/jem/article/view/295 Duvall, M.D., and Grigal, D.F. 1999. Effects of timber harvesting on coarse woody debris in red pine forests across the Great Lakes states, U.S.A. 29(12): 1926–1934. doi:10.1139/x99-158. Edmonds, Robert L., Vogt, Daniel J., Sandberg, David H., and Driver, Charles H. 1986. Decomposition of Douglas-fir and red alder wood in clear-cuttings. Canadian Journal of Forest Research 16: 822– 831. doi:10.1139/x86-145. Esseen, P.A., Ehnström, B., Ericson, L., and Sjöberg, K. 1992. Boreal forests – the focal habitats of Fennoscandia. In Ecological Principles of Nature Conservation. Edited by L. Hansson. Elsvier. pp. 252–325. Fahey, T.J., Hughes, J.W., Pu, M., and Arthur, M.A. 1988. Root decomposition and nutrient flux following whole-tree harvest of northern hardwood forest. Forest Science 34(3): 744–768. doi:10.1093/forestscience/34.3.744. Farnell, I., Elkin, C., Lilles, E., Roberts, A.-M., and Venter, M. 2020. The effects of variable retention forestry on coarse woody debris dynamics and concomitant impacts on American marten habitat after 27 years. Canadian Journal of Forest Research 50(9): 925–935. doi:10.1139/cjfr2019-0417. Fauteux, D., Imbeau, L., Drapeau, P., and Mazerolle, M.J. 2012. Small mammal responses to coarse woody debris distribution at different spatial scales in managed and unmanaged boreal forests. Forest Ecology and Management 266: 194–205. doi:10.1016/j.foreco.2011.11.020. Feller, M.C. 2003. Coarse woody debris in the old-growth forests of British Columbia. Environmental Reviews 11(S1): S135–S157. doi:10.1139/a03-008. Foster, D.R., Knight, D.H., and Franklin, J.F. 1998. Landscape patterns and legacies resulting from large, infrequent forest disturbances. Ecosystems 1(6): 497–510. doi:10.1007/s100219900046. FPPR. 2004. Forest planning and practices regulation, coarse woody debris, Part 4, Div. 5, Sec 68. BC Reg 14/2004. Available from https://www.bclaws.ca/civix/document/id/complete/statreg/14_2004#section68 [accessed 30 November 2020]. Franklin, J.F. 2019. Silvicultural principles utilized in restoration of Klamath Federated Tribes pine and mixed-conifer forests [Presentation]. North American Forest Ecology Workshop. Flagstaff, Arizona. Franklin, J.F., Shugart, H.H., and Harmon, M.E. 1987. Tree death as an ecological process. BioScience 37(8): 550–556. doi:10.2307/1310665. 64 Fraver, S., Ringvall, A., and Jonsson, B.G. 2007. Refining volume estimates of down woody debris. Canadian Journal of Forest Research 37(3): 627–633. doi:10.1139/X06-269. Fraver, S., Wagner, R.G., and Day, M. 2002. Dynamics of coarse woody debris following gap harvesting in the Acadian forest of central Maine, U.S.A. Canadian Journal of Forest Research 32(12): 2094– 2105. doi:10.1139/x02-131. Fridman, J., and Walheim, M. 2000. Amount, structure, and dynamics of dead wood on managed forestland in Sweden. Forest Ecology and Management 131(1–3): 23–36. doi:10.1016/S03781127(99)00208-X. Fuller, A.K., and Harrison, D.J. 2005. Influence of partial timber harvesting on American marten in northcentral Maine. Journal of Wildlife Management 69(2): 710–722. doi:10.2193/0022541X(2005)069[0710:IOPTHO]2.0.CO;2. Godbout, G., and Ouellet, J.-P. 2010. Fine-scale habitat selection of American marten at the southern fringe of the boreal forest. Écoscience 17(2): 175–185. doi:10.2980/17-2-3288. Gove, J.H., and Van Deusen, P.C. 2011. On fixed-area plot sampling for downed coarse woody debris. Forestry 84(2): 109–117. doi:10.1093/forestry/cpq049. Graham, R.L., and Cromack, K.J. 1982. Mass, nutrient content and decay rate of dead boles in rain forests of Olympic National Park. Canadian Journal of Forest Research 12: 511–521. doi:10.1139/x82-080. Graham, S.A. 1925. The felled tree trunk as an ecological unit. Ecology 6(4): 397–411. doi:10.2307/1929106. Greif, G.E., and Archibold, O.W. 2000. Standing-dead tree component of the boreal forest in central Saskatchewan. Forest Ecology and Management 131(1–3): 37–46. doi:10.1016/S03781127(99)00198-X. Gustafsson, L., Baker, S.C., Bauhus, J., Beese, W.J., Brodie, A., Kouki, J., Lindenmayer, D.B., Lõhmus, A., Pastur, G.M., Messier, C., Neyland, M., Palik, B., Sverdrup-Thygeson, A., Volney, W.J.A., Wayne, A., and Franklin, J.F. 2012. Retention forestry to maintain multifunctional forests: a world perspective. BioScience 62(7): 633–645. doi:10.1525/bio.2012.62.7.6. Gustafsson, L., Hannerz, M., Koivula, M., Shorohova, E., Vanha-Majamaa, I., and Weslien, J. 2020. Research on retention forestry in Northern Europe. Ecological Processes 9(1): 3. doi:10.1186/s13717-019-0208-2. Gustafsson, L., Kouki, J., and Sverdrup-Thygeson, A. 2010. Tree retention as a conservation measure in clear-cut forests of northern Europe: a review of ecological consequences. Scandinavian Journal of Forest Research 25(4): 295–308. doi:10.1080/02827581.2010.497495. Halme, P., Allen, K.A., Auniņš, A., Bradshaw, R.H.W., Brūmelis, G., Čada, V., Clear, J.L., Eriksson, A.-M., Hannon, G., Hyvärinen, E., Ikauniece, S., Iršėnaitė, R., Jonsson, B.G., Junninen, K., Kareksela, S., Komonen, A., Kotiaho, J.S., Kouki, J., Kuuluvainen, T., Mazziotta, A., Mönkkönen, M., Nyholm, K., Oldén, A., Shorohova, E., Strange, N., Toivanen, T., Vanha-Majamaa, I., Wallenius, T., Ylisirniö, A.-L., and Zin, E. 2013. Challenges of ecological restoration: lessons from forests in northern Europe. Biological Conservation 167: 248–256. doi:10.1016/j.biocon.2013.08.029. Hansen, A.J., Spies, T.A., Swanson, F.J., and Ohmann, J.L. 1991. Conserving biodiversity in managed forests. BioScience 41(6): 382–392. doi:10.2307/1311745. Harmon, Mark E., Cromack, Kermit Jr., and Smith, Bradley G. 1987. Coarse woody debris in mixedconifer forests, Sequoia National Park, California. Canadian Journal of Forest Research 17: 1265– 1272. doi:10.1139/x87-196. Harmon, M.E., and Franklin, J.F. 1989. Tree seedlings on logs in Picea-Tsuga forests of Oregon and Washington. Ecology 70(1): 48–59. doi:10.2307/1938411. Harmon, M.E., Franklin, J.F., Swanson, F.J., Gregory, S.V., Lattin, J.D., Anderson, N.H., Cline, S.P., Aumen, N.G., Sedell, J.R., Lienkaemper, G.W., and Cromack, K.Jr. 1986. Ecology of coarse woody debris in 65 temperate ecosystems. Advances in Ecological Research 15: 133–302. doi:10.1016/S00652504(08)60121-X. Heilmann-Clausen, J., and Christensen, M. 2003. Fungal diversity on decaying beech logs – implications for sustainable forestry. Biodiversity and Conservation 12: 953–973. doi:10.1023/A:1022825809503. Hély, C., Bergeron, Y., and Flannigan, M.D. 2000. Coarse woody debris in the southeastern Canadian boreal forest: composition and load variations in relation to stand replacement. Canadian Journal of Forest Research 30(5): 674–687. doi:10.1139/x99-256. Herrmann, S., Kahl, T., and Bauhus, J. 2015. Decomposition dynamics of coarse woody debris of three important central European tree species. Forest Ecosystems 2(1): 27. doi:10.1186/s40663-0150052-5. Hijmans, R.J. 2020. raster: Geographic data analysis and modeling. R package version 3.1-5. Available from https://CRAN.R-project.org/package=raster. Hodgman, T.P., Harrison, D.J., Katnik, D.D., and Elowe, K.D. 1994. Survival in an intensively trapped marten population in Maine. The Journal of Wildlife Management 58(4): 593–600. doi:10.2307/3809671. Huggard, D.J., Sandford, J., and Kremsater, L.L. 2009. Learning from habitat elements. In Forestry and biodiversity: learning how to sustain biodiversity in managed forests. Edited by F.L. Bunnell and G.B. Dunsworth. UBC Press, Vancouver, BC. pp. 152–172. Isenburg, M. 2018. LAStools: eficient LiDAR processing software. Available from https://rapidlasso.com/. Jarron, L.R. 2020. Characterization and quantification of forest secondary structure using airborne LiDAR. Doctoral dissertation, University of British Columbia, Vancouver, BC. Jia-bing, W., De-xin, G., Shi-jie, H., Mi, Z., and Chang-jie, J. 2005. Ecological functions of coarse woody debris in forest ecosystem. Journal of Forestry Research 16(3): 247–252. doi:10.1007/BF02856826. Jönsson, M.T., and Jonsson, B.G. 2007. Assessing coarse woody debris in Swedish woodland key habitats: implications for conservation and management. Forest Ecology and Management 242: 363–373. doi:10.1016/j.foreco.2007.01.054. Joyce, M.J., Erb, J.D., Sampson, B.A., and Moen, R.A. 2019. Detection of coarse woody debris using airborne light detection and ranging (LiDAR). Forest Ecology and Management 433: 678–689. doi:10.1016/j.foreco.2018.11.049. Juha, J., Kimmo, S., Sanna, S., and Olli, M. 2002. Stand structure in pine, spruce and deciduous forests: a comparison between Finnish and Russian Karelia. Scandinavian Journal of Forest Research 17: 53–61. doi:10.1080/028275802317221082. Kaytor, B. 2016. Decomposition and carbon loss in lodgepole pine (Pinus contorta var. latifolia) wood following attack by mountain pine beetle (Dendroctonous ponderosae). M.Sc Thesis, University of Northern British Columbia, Prince George, B.C. Available from https://core.ac.uk/download/pdf/84872218.pdf [accessed 10 June 2019]. Keisker, D.G. 2000. Types of wildlife trees and coarse woody debris required by wildlife of north-central British Columbia. Research Branch Ministry of Forests, Victoria, BC. Available from http://www.for.gov.bc.ca/hfd/pubs/Docs/Wp/Wp50.htm. Keren, S., and Diaci, J. 2018. Comparing the quantity and structure of deadwood in selection managed and old-growth forests in south-east Europe. Forests 9: 1–16. doi:10.3390/f9020076. Király, I., Nascimbene, J., Tinya, F., and Ódor, P. 2013. Factors influencing epiphytic bryophyte and lichen species richness at different spatial scales in managed temperate forests. Biodiversity and Conservation 22(1): 209–223. doi:10.1007/s10531-012-0415-y. Knight, H.A. 1987. The pine decline. Journal of Forestry 85(1): 25–28. 66 Koivula, M., and Vanha-Majamaa, I. 2020. Experimental evidence on biodiversity impacts of variable retention forestry, prescribed burning, and deadwood manipulation in Fennoscandia. Ecological Processes 9(1): 11. doi:10.1186/s13717-019-0209-1. Kozak, A., and Yang, R.C. 1981. Equations for estimating bark volume and thickness of commercial trees in British Columbia. Forestry chronicle 57(3): 112–115. doi:10.5558/tfc57112-3. Kuznetsova, A., Brockhoff, P.B., and Christensen, R.H.B. 2017. “lmerTest Package”: Tests in linear mixed effects models. Journal of Statistical Software 82(13): 1–26. doi:10.18637/jss.v082.i13. Laiho, R., and Prescott, C.E. 2004. Decay and nutrient dynamics of coarse woody debris in northern coniferous forests: a synthesis. Canadian Journal of Forest Research 34(4): 763–777. doi:10.1139/x03-241. Lassauce, A., Paillet, Y., Jactel, H., and Bouget, C. 2011. Deadwood as a surrogate for forest biodiversity: meta-analysis of correlations between deadwood volume and species richness of saproxylic organisms. Ecological Indicators 11: 1027–1039. doi:10.1016/j.ecolind.2011.02.004. Lenth, R. 2019. emmeans: Estimated marginal means, aka least-squares means. R package version 1.4.2. Available from https://CRAN.R-project.org/package=emmeans. Liaw, A., and Wiener, M. 2002. Classification and regression by randomForest. R News 2(3): 18–22. Available from http://CRAN.R-project.org/doc/Rnews/ Lindberg, E., Hollaus, M., Mücke, W., Fransson, J.E.S., and Pfeifer, N. 2013. Detection of lying tree stems from airborne laser scanning data using a line template matching algorithm. ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences 5: 169–174. doi:10.5194/isprsannals-II-5-W2-169-2013. Lindenmayer, D., Hobbs, R.J., Montague-Drake, R., Alexandra, J., Bennett, A., Burgman, M., Cale, P., Calhoun, A., Cramer, V., Cullen, P., Driscoll, D., Fahrig, L., Fischer, J., Franklin, J., Haila, Y., Hunter, M., Gibbons, P., Lake, S., Luck, G., MacGregor, C., McIntyre, S., Nally, R.M., Manning, A., Miller, J., Mooney, H., Noss, R., Possingham, H., Saunders, D., Schmiegelow, F., Scott, M., Simberloff, D., Sisk, T., Tabor, G., Walker, B., Wiens, J., Woinarski, J., and Zavaleta, E. 2008. A checklist for ecological management of landscapes for conservation. Ecol Letters 11(0): 78–91. doi:10.1111/j.1461-0248.2007.01114.x. Lindenmayer, D.B., and Franklin, J.F. 2002. Conserving forest biodiversity: a comprehensive multiscaled approach. Island Press. Lindenmayer, D.B., Franklin, J.F., Lõhmus, A., Baker, S.C., Bauhus, J., Beese, W., Brodie, A., Kiehl, B., Kouki, J., Pastur, G.M., Messier, C., Neyland, M., Palik, B., Sverdrup-Thygeson, A., Volney, J., Wayne, A., and Gustafsson, L. 2012. A major shift to the retention approach for forestry can help resolve some global forest sustainability issues: retention forestry for sustainable forests. Conservation Letters 5(6): 421–431. doi:10.1111/j.1755-263X.2012.00257.x. Lofroth, E.C. 1993. Scale dependent analysis of habitat selection by marten in the Sub-Boreal Spruce Biogeoclimatic Zone, British Columbia. MSc thesis. Simon Fraser University, Burnaby, B.C. Lohr, S.M., Gauthreaux, S.A., and Kilgo, J.C. 2002. Importance of coarse woody debris to avian communities in loblolly pine forests. Conservation Biology 16(3): 767–777. doi:10.1046/j.15231739.2002.01019.x. Magnússon, R.Í., Tietema, A., Cornelissen, J.H.C., Hefting, M.M., and Kalbitz, K. 2016. Tamm review: sequestration of carbon from coarse woody debris in forest soils. Forest Ecology and Management 377: 1–15. doi:10.1016/j.foreco.2016.06.033. Martinuzzi, S., Vierling, L.A., Gould, W.A., Falkowski, M.J., Evans, J.S., Hudak, A.T., and Vierling, K.T. 2009. Mapping snags and understory shrubs for a LiDAR-based assessment of wildlife habitat suitability. Remote Sensing of Environment 113(12): 2533–2546. doi:10.1016/j.rse.2009.07.002. Maser, C., Tarrant, R.F., Trappe, J.M., and Franklin, J.F. 1988. From the forest to the sea: a story of fallen trees. USDA Forest Service general technical report PNW-GTR-229. 67 McCarthy, B.C., and Bailey, R.R. 1994. Distribution and abundance of coarse woody debris in a managed forest landscape of the central Appalachians. Canadian Journal of Forest Research 24(7): 1317– 1329. doi:10.1139/x94-172. McCarthy, B.C., Small, C.J., and Rubino, D.L. 2001. Composition, structure and dynamics of Dysart Woods, an old-growth mixed mesophytic forest of southeastern Ohio. Forest Ecology and Management 140: 193–213. doi:10.1016/S0378-1127(00)00280-2. McGee, G.G., and Kimmerer, R.W. 2002. Forest age and management effects on epiphytic bryophyte communities in Adirondack northern hardwood forests, New York, U.S.A. Canadian Journal of Forest Research 32: 15. doi:10.1139/x02-083 Mclaren, B.E., Gammond, P.R., and Thompson, I.D. 2013. Characteristics of the boreal mixedwood forest associated with the use of subnivean access points by American martens. Écoscience 20(4): 383– 390. doi:10.2980/20-4-3634. McMinn, J.W., and Crossley, D.A. 1996. Biodiversity and coarse woody debris in southern forests. DIANE Publishing. Meidinger, D.V., and Pojar, J. 1991. Ecosystems of British Columbia. Special Report Series 6, BC Ministry of Forest, Victoria, BC. Minnich, C., Peršoh, D., Poll, C., and Borken, W. 2020. Changes in chemical and microbial soil parameters following 8 years of deadwood decay: an experiment with logs of 13 tree species in 30 forests. Ecosystems. doi:10.1007/s10021-020-00562-z. Moroni, M.T. 2006. Disturbance history affects dead wood abundance in Newfoundland boreal forests. Canadian Journal of Forest Research 36(12): 3194–3208. doi:10.1139/x06-195. Morrissey, R.C., Jenkins, M.A., and Saunders, M.R. 2014. Accumulation and connectivity of coarse woody debris in partial harvest and unmanaged relict forests. PLoS ONE 9(11): e113323. doi:10.1371/journal.pone.0113323. Mücke, W., Deák, B., Schroiff, A., and Hollaus, M. 2013. Detection of fallen trees in forested areas using small footprint airborne laser scanning data. Canadian Journal of Remote Sensing 39: S32–S40. doi:10.5589/m13-013. Nigh, G. 2016. Total and merchantable volume equations for common tree species in British Columbia: by region and biogeoclimatic zone. Prov. B.C. Technical Report 106, Victoria, B.C. Available from www.for.gov.bc.ca/hfd/pubs/Docs/Tr/Tr106.html. Nyström, M., Holmgren, J., Fransson, J.E.S., and Olsson, H. 2014. Detection of windthrown trees using airborne laser scanning. International Journal of Applied Earth Observation and Geoinformation 30: 21–29. doi:10.1016/j.jag.2014.01.012. Ódor, P., Heilmann-Clausen, J., Christensen, M., Aude, E., van Dort, K.W., Piltaver, A., Siller, I., Veerkamp, M.T., Walleyn, R., Standovár, T., van Hees, A.F.M., Kosec, J., Matočec, N., Kraigher, H., and Grebenc, T. 2006. Diversity of dead wood inhabiting fungi and bryophytes in semi-natural beech forests in Europe. Biological Conservation 131(1): 58–71. doi:10.1016/j.biocon.2006.02.004. Payer, D., and Harrison, D. 2005. Relationships between forest structure and habitat use by American martens in Maine, USA. In Martens and fishers (Martes) in human-altered environments: and international perspective. Edited by D.J. Harrison, A.K. Fuller, and G. Proulx. Springer, New York. pp. 173–186. doi:10.1007/0-387-22691-5_8. Payer, D.C. 1999. Influences of timber harvesting and trapping on habitat selection and demographic characteristics of American marten. Ph.D. dissertation, University of Maine, Orono, Maine, USA. Payer, D.C., and Harrison, D.J. 2000. Structural differences between forests regenerating following spruce budworm defoliation and clear-cut harvesting: implications for marten. 30(12): 1965– 1972. doi:10.1139/x00-129. 68 Payer, D.C., and Harrison, D.J. 2003. Influence of forest structure on habitat use by American marten in an industrial forest. Forest Ecology and Management 179(1–3): 145–156. doi:10.1016/S03781127(02)00517-0. Pedlar, J.H., Pearce, J.L., Venier, L.A., and McKenney, D.W. 2002. Coarse woody debris in relation to disturbance and forest type in boreal Canada. Forest Ecology and Management 158(1): 189–194. doi:10.1016/S0378-1127(00)00711-8. Pesonen, A., Maltamo, M., Eerikäinen, K., and Packalèn, P. 2008. Airborne laser scanning-based prediction of coarse woody debris volumes in a conservation area. Forest Ecology and Management 255(8): 3288–3296. doi:10.1016/j.foreco.2008.02.017. Pinheiro, J., Bates, D., DebRoy, S., and Sarkar, D. 2016. nlme: linear and nonlinear mixed effects models. R package version 3.1-118. Polley, E., LeDell, E., Kennedy, C., and van der Laan, M. 2019. SuperLearner: Super Learner prediction. R package version 2.0-26. Available from https://CRAN.R-project.org/package=SuperLearner. Poole, K.G., Porter, A.D., deVries, A., Maundrell, C., Grindal, S.D., and St. Clair, C.C. 2004. Suitability of a young deciduous-dominated forest for American marten and the effect of forest removal. Canadian Journal of Zoology 82(3): 423–435. doi:10.1139/z04-006. Porter, A.D., St Clair, C.C., and Vries, A. de. 2005. Fine-scale selection by marten during winter in a young deciduous forest. Canadian Journal of Forest Research 35(4): 901–909. doi:10.1139/x05-014. Potvin, F., Bélanger, L., and Lowell, K. 2000. Marten habitat selection in a clearcut Boreal landscape. Conservation Biology 14(3): 844–857. doi:10.1046/j.1523-1739.2000.99036.x. Proulx, G., and Aubry, K.B. 2020. The Martes complex: opportunities for developing multi-species management and conservation strategies. Canadian Wildlife Biology and Management 9(2): 49– 67. Queiroz, G.L., McDermid, G., Linke, J., Hopkinson, C., and Kariyeva, J. 2020. Estimating coarse woody debris volume using image analysis and multispectral LiDAR. Forests 11(2): 141. doi:10.3390/f11020141. R Core Team. 2020. A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available from https://www.R-project.org/. Rambo, T. 2001. Decaying logs and habitat heterogeneity: implications for bryophyte diversity in Western Oregon forests. Northwest Science 75: 270–277. Available from http://hdl.handle.net/2376/940 Robertson, P.A., and Bowser, Y.H. 1999. Coarse woody debris in mature Pinus ponderosa stands in Colorado. Journal of the Torrey Botanical Society 126(3): 255–267. doi:10.2307/2997280. Rouvinen, S., Kuuluvainen, T., and Karjalainen, L. 2002. Coarse woody debris in old Pinus sylvestris dominated forests along a geographic and human impact gradient in boreal Fennoscandia. Canadian Journal of Forest Research 32(12): 2184–2200. doi:10.1139/x02-144. Rubino, D.L., and McCarthy, B.C. 2003. Evaluation of coarse woody debris and forest vegetation across topographic gradients in a southern Ohio forest. Forest Ecology and Management 183: 221–238. doi:10.1016/S0378-1127(03)00108-7. Sandström, J., Bernes, C., Junninen, K., Lõhmus, A., Macdonald, E., Müller, J., and Jonsson, B.G. 2019. Impacts of dead wood manipulation on the biodiversity of temperate and boreal forests. A systematic review. Journal of Applied Ecology 56(7): 1770–1781. doi:10.1111/1365-2664.13395. Schiegg, K. 2000. Effects of dead wood volume and connectivity on saproxylic insect species diversity. Écoscience 7(3): 290–298. doi:10.1080/11956860.2000.11682598. Schütt, P., and Cowling, E.B. 1985. Waldsterben, a general decline of forests in central Europe: symptoms, development and possible causes. Plant Disease 69(7): 548–558. Available from https://www.cabdirect.org/cabdirect/abstract/19861314547 69 Seip, C.R., Hodder, D.P., Crowley, S.M., and Johnson, C.J. 2018. Use of constructed coarse woody debris corridors in a clearcut by American martens (Martes americana) and their prey. Forestry: An International Journal of Forest Research 91(4): 506–513. doi:10.1093/forestry/cpy010. Short, L.E., Gabet, E.J., and Hoffman, D.F. 2015. The role of large woody debris in modulating the dispersal of a post-fire sediment pulse. Geomorphology 246: 351–358. doi:10.1016/j.geomorph.2015.06.031. Siccama, T.G., Bliss, M., and Vogelmann, H.W. 1982. Decline of red spruce in the green mountains of vermont. Bulletin of the Torrey Botanical Club 109(2): 162. doi:10.2307/2996256. Siitonen, J. 2001. Forest management, coarse woody debris and saproxylic organisms: Fennoscandian boreal forests as an example. Ecological bulletins: 11–41. Available from https://www.jstor.org/stable/20113262 Siitonen, J., and Martikainen, P. 1994. Occurrence of rare and threatened insects living on decaying Populus Tremula: a comparison between Finnish and Russian Karelia. Scandinavian Journal of Forest Research 9: 185–191. doi:10.1080/02827589409382830. Siitonen, J., Martikainen, P., Punttila, P., and Rauh, J. 2000. Coarse woody debris and stand characteristics in mature managed and old-growth boreal mesic forests in southern Finland. Forest Ecology and Management 128(3): 211–225. doi:10.1016/S0378-1127(99)00148-6. Sippola, A., Siitonen, J., and Kallio, R. 1998. Amount and quality of coarse woody debris in natural and managed coniferous forests near the timberline in Finnish Lapland. Scandinavian Journal of Forest Research 13(1–4): 204–214. doi:10.1080/02827589809382978. Smith, A.C., and Schaefer, J.A. 2002. Home-range size and habitat selection by American marten ( Martes americana ) in Labrador. Canadian Journal of Zoology 80: 1602–1609. doi:10.1139/z02166. Sollins, P., Cline, S.P., Verhoeven, T., Sachs, D., and Spycher, G. 1987. Patterns of log decay in old-growth Douglas-fir forests. Canadian Journal of Forest Research 17(12): 1585–1595. doi:10.1139/x87243 Spears, J.D., Holub, S.M., Harmon, M.E., and Lajtha, K. 2003. The influence of decomposing logs on soil biology and nutrient cycling in an old-growth mixed coniferous forest in Oregon, U.S.A. Canadian Journal of Forest Research 33(11): 2193–2201. doi:10.1139/x03-148. Spencer, W.D., Barrett, R.H., and Zielinski, W.J. 1983. Marten habitat preferences in the northern Sierra Nevada. The Journal of Wildlife Management 47(4): 1181–1186. doi:10.2307/3808189. Spies, T.A., and Cline, S.P. 1988. Coarse woody debris in forests and plantations of coastal Oregon. In From the forest to the sea: a story of fallen trees. Pacific Northwest Research Station, Forest Service, US Department of Agriculture, Portland, OR. pp. 5–23. Spies, T.A., Franklin, J.F., and Thomas, T.B. 1988. Coarse woody debris in Douglas-fir forests of western Oregon and Washington. Ecology 69(6): 1689–1702. doi:10.2307/1941147. Spribille, T., Thor, G., Bunnell, F.L., Goward, T., and Björk, C.R. 2008. Lichens on dead wood: speciessubstrate relationships in the epiphytic lichen floras of the Pacific Northwest and Fennoscandia. Ecography 31(6): 741–750. doi:10.1111/j.1600-0587.2008.05503.x. Stevenson, S.K., Jull, M.J., and Rogers, B.J. 2006. Abundance and attributes of wildlife trees and coarse woody debris at three silvicultural systems study areas in the Interior Cedar-Hemlock Zone, British Columbia. Forest Ecology and Management 233(1): 176–191. doi:10.1016/j.foreco.2006.06.023. Stevens-Rumann, C., Morgan, P., and Hoffman, C. 2015. Bark beetles and wildfires: How does forest recovery change with repeated disturbances in mixed conifer forests? Ecosphere 6(6): 1–17. doi:10.1890/ES14-00443.1. Steventon, D. 2014. Recommendations for a multi-scale habitat analysis for the marten (Martes americana). Ministry of Forests, Lands, Natural Resources and Rural Development. 70 Stokland, J., Siitonen, J., and Jonsson, B.G. 2012. Biodiversity in dead wood. Cambridge University Press, Cambridge, UK. Stokland, J.N. 2001. The coarse woody debris profile: an archive of recent forest history and an important biodiversity indicator. Ecological Bulletins 49: 71–83. Available from https://www.jstor.org/stable/20113265 Stone, J., Parminter, J., Arsenault, A., Manning, T., Densmore, N., Davis, G., and MacKinnon, A. 2002. Dead tree management in British Columbia. In: Laudenslayer, William F., Jr.; Shea, Patrick J.; Valentine, Bradley E.; Weatherspoon, C. Phillip; Lisle, Thomas E., technical coordinators. Proceedings of the symposium on the ecology and management of dead wood in western forests. 1999 November 2-4; Reno, NV. Gen. Tech. Rep. PSW-GTR-181. Albany, CA: Pacific Southwest Research Station, Forest Service, U.S. Department of Agriculture. Stone, J.N., MacKinnon, A., Parminter, J.V., and Lertzman, K.P. 1998. Coarse woody debris decomposition documented over 65 years on southern Vancouver Island. Canadian Journal of Forest Research 28: 788–793. doi:10.1139/x98-047. Storch, F., Dormann, C.F., and Bauhus, J. 2018. Quantifying forest structural diversity based on largescale inventory data: a new approach to support biodiversity monitoring. For. Ecosyst. 5(1): 34. doi:10.1186/s40663-018-0151-1. Sturtevant, B.R., Bissonette, J.A., and Long, J.N. 1996. Temporal and spatial dynamics of boreal forest structure in western Newfoundland: silvicultural implications for marten habitat management. Forest Ecology and Management 87: 13–25. doi:10.1016/S0378-1127(96)03837-6. Sturtevant, B.R., Bissonette, J.A., Long, J.N., and Roberts, D.W. 1997. Coarse woody debris as a function of age, stand structure, and disturbance in boreal Newfoundland. Ecological Applications 7(2): 702–712. doi:10.2307/2269532. Sullivan, T.P., and Sullivan, D.S. 2019. Long-term functionality of woody debris structures for forest-floor small mammals on clearcuts. Forest Ecology and Management 451: 117535. doi:10.1016/j.foreco.2019.117535. Sullivan, T.P., and Sullivan, D.S. 2020. Similarity in occupancy of different-sized forest patches by small mammals on clearcuts: conservation implications for red-backed voles and small mustelids. Mammal Research 65(2): 255–266. doi:10.1007/s13364-019-00467-w. Sullivan, T.P., Sullivan, D.S., and Sullivan, J.H. 2017. Mammalian responses to windrows of woody debris on clearcuts: abundance and diversity of forest-floor small mammals and presence of small mustelids. Forest Ecology and Management 399: 143–154. doi:10.1016/j.foreco.2017.05.028. Sumnall, M.J., Hill, R.A., and Hinsley, S.A. 2016. Comparison of small-footprint discrete return and full waveform airborne lidar data for estimating multiple forest variables. Remote Sensing of Environment 173: 214–223. doi:10.1016/j.rse.2015.07.027. Taboada, Á., Tárrega, R., Calvo, L., Marcos, E., Marcos, J.A., and Salgado, J.M. 2010. Plant and carabid beetle species diversity in relation to forest type and structural heterogeneity. European Journal of Forest Research 129(1): 31–45. doi:10.1007/s10342-008-0245-3. Thompson, I.D., Baker, J.A., Jastrebski, C., Dacosta, J., Fryxell, J., and Corbett, D. 2008. Effects of postharvest silviculture on use of boreal forest stands by amphibians and marten in Ontario. The Forestry Chronicle 84(5): 741–747. doi:10.5558/tfc84741-5. Thompson, I.D., Fryxell, J.M., and Harrison, D.J. 2012. Improved insights into use of habitat by American martens. In Biology and conservation of martens, sabes, and fishers. Edited by K.B. Aubry, W.J. Zielinski, M.J. Raphael, G. Proulx, and S.W. Buskirk. Cornell University Press, Ithaca, New York. pp. 209–230. Thorn, S., Seibold, S., Leverkus, A.B., Michler, T., Müller, J., Noss, R.F., Stork, N., Vogel, S., and Lindenmayer, D.B. 2020. The living dead: acknowledging life after tree death to stop forest degradation. Front. Ecol. Environ. 18(9): 505–512. doi:10.1002/fee.2252. 71 Van Wagner, C.E. 1968. The line intersect method in forest fuel sampling. Forest Science 14(1): 20–26. doi:10.1093/forestscience/14.1.20. Vanderwel, M.C., Coomes, D.A., and Purves, D.W. 2013. Quantifying variation in forest disturbance, and its effects on aboveground biomass dynamics, across the eastern United States. Global Change Biology 19(5): 1504–1517. doi:10.1111/gcb.12152. Vanderwel, M.C., Thorpe, H.C., Shuter, J.L., Caspersen, J.P., and Thomas, S.C. 2008. Contrasting downed woody debris dynamics in managed and unmanaged northern hardwood stands. Canadian Journal of Forest Research 38(11): 2850–2861. doi:10.1139/X08-130. Wang, S., Knapp, B.O., Ehlers, S., Graham, B., Gao, X., and Timm, S. 2021. Forest management effects on downed dead wood at stand and landscape scales in a temperate forest of the central United States. Forest Ecology and Management 482: 1–12. doi:10.1016/j.foreco.2020.118905. Wang, T., Hamann, A., Spittlehouse, D.L., and Carroll, C. 2016. Locally downscaled and spatially customizable climate data for historical and future periods for North America. PLoS ONE 11(6): e0156720. doi:10.1371/journal.pone.0156720. Warren, G. R. and Meades, J. P. 1986. Wood defect and density studies II: total and net volume equations for Newfoundland’s forest management units. Can. For. Serv. Newfoundland For. Cent. Inf. Rep. N-X-242. White, J.C., Wulder, M.A., Varhola, A., Vastaranta, M., Coops, N.C., Cook, B.D., Pitt, D., and Woods, M. 2013. A best practices guide for generating forest inventory attributes from airborne laser scanning data using an area-based approach. The Forestry Chronicle 89(06): 722–723. doi:10.5558/tfc2013-132. White, S.D., Hart, J.L., Schweitzer, C.J., and Dey, D.C. 2015. Altered structural development and accelerated succession from intermediate-scale wind disturbance in Quercus stands on the Cumberland Plateau, USA. Forest Ecology and Management 336: 52–64. doi:10.1016/j.foreco.2014.10.013. Wiebe, P.A., Thompson, I.D., McKague, C.I., Fryxell, J.M., and Baker, J.A. 2014. Fine-scale winter resource selection by American martens in boreal forests and the effect of snow depth on access to coarse woody debris. Ecoscience 21(2): 123–132. doi:10.2980/21-2-3687. Wing, B.M., Ritchie, M.W., Boston, K., Cohen, W.B., Gitelman, A., and Olsen, M.J. 2012. Prediction of understory vegetation cover with airborne lidar in an interior ponderosa pine forest. Remote Sensing of Environment 124: 730–741. doi:10.1016/j.rse.2012.06.024. Wright, M.N., and Ziegler, A. 2017. ranger: A fast implementation of Random Forests for high dimensional data in C++ and R. J. Stat. Soft. 77(1): 1–17. doi:10.18637/jss.v077.i01. Ylisirniö, A.-L., Berglund, H., Aakala, T., Kuuluvainen, T., Kuparinen, A.-M., Norokorpi, Y., Hallikainen, V., Mikkola, K., and Huhta, E. 2009. Spatial distribution of dead wood and the occurrence of five saproxylic fungi in old-growth timberline spruce forests in northern Finland. Scandinavian Journal of Forest Research 24(6): 527–540. doi:10.1080/02827580903362489. 72